Evaluation of the University of Minnesota Easy Culture System II and the

3M Petrifilm for diagnosis of mastitis causing organisms

A Thesis

Submitted to the Graduate Faculty

in Partial Fulfillment of the Requirements

for the Degree of

Master of Science

in the Department of Health Management

Atlantic Veterinary College

University of Prince Edward Island

Jennifer Lea McCarron

Charlottetown, P. E. I.

2012

C 2012. J. McCarron Library and Archives Bibliotheque et Canada Archives Canada

Published Heritage Direction du 1+1 Branch Patrimoine de I'edition 395 Wellington Street 395, rue Wellington Ottawa ON K1A0N4 Ottawa ON K1A 0N4 Canada Canada Your file Votre reference ISBN: 978-0-494-94066-2

Our file Notre reference ISBN: 978-0-494-94066-2

NOTICE: AVIS: The author has granted a non­ L'auteur a accorde une licence non exclusive exclusive license allowing Library and permettant a la Bibliotheque et Archives Archives Canada to reproduce, Canada de reproduire, publier, archiver, publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public communicate to the public by par telecommunication ou par I'lnternet, preter, telecommunication or on the Internet, distribuer et vendre des theses partout dans le loan, distrbute and sell theses monde, a des fins commerciales ou autres, sur worldwide, for commercial or non­ support microforme, papier, electronique et/ou commercial purposes, in microform, autres formats. paper, electronic and/or any other formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur ownership and moral rights in this et des droits moraux qui protege cette these. Ni thesis. Neither the thesis nor la these ni des extraits substantiels de celle-ci substantial extracts from it may be ne doivent etre imprimes ou autrement printed or otherwise reproduced reproduits sans son autorisation. without the author's permission.

In compliance with the Canadian Conformement a la loi canadienne sur la Privacy Act some supporting forms protection de la vie privee, quelques may have been removed from this formulaires secondaires ont ete enleves de thesis. cette these.

While these forms may be included Bien que ces formulaires aient inclus dans in the document page count, their la pagination, il n'y aura aucun contenu removal does not represent any loss manquant. of content from the thesis. Canada CONDITION OF USE

The author has agreed that the Library, University of Prince Edward Island, may make this thesis freely available for inspection. Moreover, the author has agreed that permission for extensive copying of this thesis for scholarly purposes may be granted by the professor or professors who supervised the thesis work recorded herein or, in their absence, by the Chairman of the Department or the Dean of the Faculty in which the thesis work was done. It is understood that due recognition will be given to the author of this thesis and to the University of Prince Edward Island in any use of the material in this thesis. Copying or publication or any other use of the thesis for financial gain without approval by the University of Prince Edward Island and the author’s written permission is prohibited.

Requests for permission to copy or to make any other use of material in this thesis in whole or in part should be addressed to:

Chairman of the Department of Health management

Atlantic Veterinary College

University of Prince Edward Island

Charlottetown, P. E. I.

Canada CIA 4P3 SIGNATURE PAGE

da)*Civ)

REMOVED Abstract

Despite extensive research in the field of bovine mastitis, it is still widely considered the most costly infectious disease of dairy cattle. In Canada, mastitis treatment accounts for more than half of all antibiotics used by dairy producers. Public concern over the use of such antibiotics in animal agriculture requires the prudent use of these products on dairy farms. In the mid-1990s, researchers began to call for a more targeted approach to mastitis therapy. Several studies have been conducted to find appropriate diagnostic tools that could be used in such a manner. However, further research is required to validate the tests that are currently available. The objective of the first study was to determine test characteristics (versus standard culture techniques) of 2 potential on-farm culture systems for clinical mastitis, the Minnesota Easy Culture System II Bi-plate (University of Minnesota, Minneapolis, MN) and the Petrifilm (3M, Minneapolis, MN). The tests were evaluated using 282 clinically positive mastitic milk samples (not frozen) from 21 commercial dairy farms to determine their ability to differentiate appropriate treatment groups. All cases that had gram positive growth were considered treatment candidates, while cases that grew gram negative organisms only, or yielded no bacterial growth, were classified as no treatment. Both the Bi-plate and 3M Petrifilm systems were highly sensitive, 97.9% (95% Cl: 94.0 - 99.6) and 93.8% (95% Cl: 87.7 - 97.5), respectively. The very high negative predictive values (96.4% (95% Cl: 89.9 - 99.3) for Bi-plate and 89.4% (95% Cl: 79.4 - 95.6) for Petrifilm) were important attributes of each of these tests, which will minimize the number of cases requiring therapy that go untreated. The results indicate that both tests have the potential to appropriately categorize clinically mastitic cases. The primary objective of the second study was to compare microbiological results of the University of Minnesota Tri-plate and the 3M Petrifilm Staph Express (STX) Count Plate to standard culture techniques for identification of clinical mastitis causedS. aureus.by The secondary objective of this study was to evaluate the Tri-plates ability to differentiate Streptococcus spp. from other gram positive organisms. Both the University of Minnesota Tri-plate and the 3M STX Petrifilm were able to successfully detectS. aureus in 271 clinically positive mastitic milk samples (not frozen) from 21 commercial dairy farms when the tests were used in a laboratory setting. The presence of P hemolysis on the Tri-plate was a highly sensitive method (97.9%, 95% Cl: 88.7 - 99.0) to diagnoseS. aureus when read by a trained laboratory technician. Yet, when read by individuals with limited microbiology experience, the Se of the Tri-plate was much lower, ranging from 43.2% to 59.1%. Specificity of the Tri-plate was higher when read by the inexperienced readers (93.8% - 95.9%) than by the laboratory technician (81.8%). The Tri-plate was able to successfully differentiateStrep, spp. from other gram positive organisms in clinically mastitic cases. Finally, the objective of the third study was to evaluate the use of two on-farm tests as part of a selective dry cow therapy program. The 3M Petrifilm and the University of Minnesota Tri-plate were evaluated on their ability to classify a sample as either infected or uninfected at the time of dry-off. Each test was evaluated individually as well as in combination with somatic cell count data obtained from monthly Dairy Herd Improvement records. Using the 3M Petrifilm alone to classify a sample as infected resulted in a

v sensitivity of 100% (95% Cl: 94.7 - 100) and a specificity of 61.2% (95% Cl: 53.8 - 68.1). Using the Tri-plate alone to classify a sample as infected resulted in a sensitivity of 96.4% (95% Cl: 90.2 - 98.8) and a specificity of 44.6% (95% CT. 35.9 - 53.6). Both the 3M Petrifilm and University of Minnesota Tri-plate are rapid and reliable tests that could be used successfully in a selective dry cow therapy program. ACKNOWLEDGEMENTS

This research was financed by NSERC, Alberta Milk, Dairy Farmers of New Brunswick, Nova Scotia, Ontario and Prince Edward Island, Novalait Inc., Dairy Farmers of Canada, Canadian Dairy Network, AAFC, PHAC, Technology PEI Inc., University de Montreal and University of Prince Edward Island through the Canadian Bovine Mastitis Research Network.

I would like to thank everyone who was involved in the completion of this thesis. I feel fortunate to have worked with many supportive faculty and staff at the Atlantic Veterinary College.

Thank you to the members of my supervisory committee Drs. Greg Keefe, Shawn McKenna, Ian Dohoo, Anne Muckle and Henrik Stryhn.

Thank you to the laboratory staff members who taught me so much and helped make this project possible, Doris Poole, Shana Richard, Zoe Little.

I would also like to acknowledge my co-workers in Farm Service who not only helped a great deal with this project but are great Mends as well, Theresa Andrews, Lloyd Dalziel and Ricky Milton.

Special thanks go to Dr. Shawn McKenna, an excellent mentor who helped keep me motivated and was always a great Mend throughout the duration of my time at AVC.

Most importantly I would like to thank my husband Ryan, my best friend who makes every day an adventure. And finally I would like to thank my children, Ellen and Alexander. Although they were not present for the majority of this thesis they definitely had a part to play in its completion... by seriously delaying it! Table of Contents

Table of Contents...... viii List of Abbreviations...... x List o f Tables...... 1 List of Figures...... 2 CHAPTER 1. INTRODUCTION TO MASTITIS DIAGNOSTICS AND THE NEED FOR ON-FARM PATHOGEN IDENTIFICATION SYSTEMS...... 3 1.1. Introduction...... 4 1.2. Microbiologic Tests for Mastitis...... 8 1.3. Study Objectives...... 11 1.4. References...... 12 CHAPTER 2 . LABORATORY EVLAUATION OF 3M PETRIFILM AND UNIVERSITY OF MINNESOTA BI-PLATES AS POTENTIAL ON-FARM TESTS FOR CLINICAL MASTITIS...... 15 2.1. Abstract...... 16 2.2. Introduction...... 17 2.3. Material and methods...... 19 2.4. Results...... 24 2.5. Discussion...... 28 2.6. Conclusion...... 35 2.7. Acknowledgements...... 35 2.8. Sources and manufacturers...... 36 2.9. References...... 36 CHAPTER 3 . EVALUATION OF THE UNIVERSITY OF MINNESOTA TRI-PLATE AND 3M PETRIFILM FOR THE ISOLATION OF STAPHYLOCOCCUS AUREUS AND STREPTOCOCCUS SPP. FROM CLINICALLY MASTITC MILK SAMPLES...... 44 3.1. Abstract...... 45 3.2. Introduction...... 46 3.3. Materials and Methods...... 48 3.4. Results...... 51 3.5. Discussion...... 55 3.6. Conclusion...... 61 3.8. References...... 62 CHAPTER 4. EVALUATION OF 3M PETRIFILM, UNIVERSITY OF MINNESOTA TRI­ PLATE AND SOMATIC CELL COUNT AS DIAGNOSTIC TESTS USED TO DETERMINE INFECTION STATUS AT TIME OF DRY-OFF...... 70 4.1. Abstract...... 71 4.2. Introduction...... 72 4.3. Materials and Methods...... 74 4.4. Results...... 79 4.5. Discussion...... 83 4.6. Conclusions...... 89 4.7. Acknowledgments...... 90 4.8. References...... 90 CHAPTER 5. SUMMARY AND GENERAL DISCUSSION ...... 98 5.1. Introduction...... 99 5.2. Minnesota Easy Culture System...... II 100 5.3. 3M Petrifilm...... 103 5.4. Use in Selective Dry Cow Therapy...... 105 5.5. Overall Conclusions...... 106 5.6. Future Directions...... 107 5.7. References...... 109 APPENDIX 1: NMC GUIDELINES OF SIGNIFICANCE...... 110 List of Abbreviations

AC Aerobic Count CBMRN Canadian Bovine Mastitis Research Network CC Coliform Count DHI Dairy Herd Improvement GEE Generalized Estimating Equations IMI Intramammary Infection ITS Internal Teat Sealant MTKT Modified thallium sulfate-crystal violet - B toxin blood NMC National Mastitis Council NPV Negative Predictive Value PPV Positive Predictive Value Se Sensitivity see Somatic Cell Count Sp Specificity STX Staph Express

X List of Tables

Table 2.1.Gold standard microbiology results of samples submitted to be evaluated by Bi­ plates and Petrifilms...... 39 Table2.2. Agreement between technician and automated Petrifilm reader for Aerobic Count and Coliform Count Petrifilms using diluted milk samples...... 40 Table 2.3.Test characteristics of the Petrifilm test system that uses a threshold of <20 colonies on the Coliform Count plate and >5 colonies on the Aerobic Count plate, based on its ability to correctly classify a sample as gram positive...... 41 Table 2.4.Test characteristics of the University of Minnesota Bi-plate, based on its ability to correctly classify samples as gram positive ...... 42 Table 3.1.Gold standard microbiology results of samples used to evaluate the University of Minnesota Tri-plate and the 3M Petrifilm Staph Express...... 64 Table 3.2.Test characteristics of the University of Minnesota Tri-plate when read by an experienced technician, based on its ability to correctly classify samples as S. aureus...... 65 Table 3.3.Test characteristics of the University of Minnesota Tri-plate based on its ability to correctly classify samples as S. aureus when read by masked readers with limited microbiology training...... 66 Table 3.4.Test characteristics of the University of Minnesota Tri-plate when read by an experienced technician, based on its ability to correctly classify samples as Strep, spp...... 67 Table 3.5.Test characteristics of the 3M Staph Express Petrifilm when read by an experienced technician using the presence of red-violet colonies on initial culture to classify samples as S. aureus vs. using the confirmatory disk...... 68 Table 4.1.Criteria used to create definitions for test positive samples...... 92 Table 4.2.Sensitivity and specificity of all test definitions evaluated versus gold standard status, based on NMC 1987 significance...... 93

1 List of Figures

Figure 2.1. Comparison of sensitivity and specificity for various combinations of colony count thresholds (grouped as 5,10 and 20 colony forming units respectively) for the diluted Aerobic Count (AC) and Coliform Count Petrifilms...... 43 Figure 4.1. Gold standard culture results...... 94 Figure 4.2. Positive predictive values for a range of disease prevalences...... 95 Figure 4.3. Negative predictive values for a range of disease prevalences...... 96

2 CHAPTER 1. INTRODUCTION TO MASTITIS DIAGNOSTICS AND THE NEED FOR ON-FARM PATHOGEN IDENTIFICATION SYSTEMS

3 1.1. Introduction

Despite extensive research in the field of bovine mastitis, it is still generally considered the most costly infectious disease of dairy cattle (1). Costs associated with the disease can be attributed to reductions in milk yield, discarded milk, cost of therapy, veterinary care, and culling or death of affected animals (2). A recent Canadian study of clinical mastitis on dairy farms found the mean incidence rate of clinical mastitis to be 23 cases per 100 cow years (3).

In Canada, mastitis treatment accounts for more than half of all antibiotics used by dairy producers (4). Public concern over the use of such antibiotics in animal agriculture requires the prudent use of these products on dairy farms. Concerns that their use may promote bacterial antibiotic resistance (5) and leave residues in the food chain have also been identified. The practice of using antibiotics to treat clinical cases of mastitis without knowledge of the causative organism is widespread on Canadian dairy farms. However, a considerable proportion of the antibiotic used to treat these cases may not be justified because of high self-cure rates and poor efficacy of treatments against certain pathogens. For example, the use of intramammary antimicrobials for the treatment ofEscherichia coli clinical cases was most often ineffective due to the short duration of infections and high spontaneous cure rates (6,7). In 2008, Olde Riekerinket al. reported that 43.9% of 3033 milk samples submitted from Canadian cows with clinical mastitis yielded no bacterial growth (3). Negative culture results may be due to spontaneous clearance of pathogens, cyclical shedding ofStaphylococcus aureus, phagocytosis of bacteria within the milk sample and the presence of too few bacterial colonies to yield a diagnosis (8).

4 In the mid-1990s, researchers began to call for a more targeted approach to mastitis therapy. The need for an appropriate method to identify cases of mastitis in which antibiotic use would be justified was also identified (9). Several studies have been conducted to find appropriate diagnostic tools that could be used in such a manner. In 2002, a review completed by Leslie et al. evaluated various cow-side products that have been developed to improve management practices on dairy farms (10). The authors concluded that further research was required to improve the tests that were currently available and to create novel products that would supply dairy producers with economical, rapid and easy to use cow-side methods of testing.

Selective growth media may be useful on-farm tools for determining the major pathogen categories that cause mastitis, thus enabling a producer to choose an appropriate treatment. Many large dairy farms in the US have established milk culturing laboratories and there has been a perceived increase in the adoption of on-farm culture systems for the selective treatment of clinical mastitis (11). In Canada, the demographics of the dairy industry are different than in the US. The average dairy herd size in Canada in 2007 was 68 cows (12), whereas, the average herd size in the US in 2007 was 125 cows (13). The size of a herd may play a role in the uptake and type of on-farm test system that is suitable. The results generated from these on-farm laboratories could be used to develop treatment protocols and monitor long term mastitis situations (11). Goddenet al. reiterated the need for the adoption of rapid on-farm culture systems that would allow producers to make informed treatment decisions (14). By treating fewer cases with antibiotics and discarding less milk that contains residues, the cost per case of mastitis was reduced (15). Tests that

5 will play a part in reducing the economic impacts of mastitis, as well as promote the prudent use of antimicrobials on dairy farms, need to be validated (14).

On-farm tests that are rapid, sensitive and have high negative predictive values would be valuable tools for producers. A targeted approach to mastitis therapy could result in the reduction in the amount of antibiotics used as cows not requiring antibiotic therapy could be differentiated from those that do. In the creation of an on-farm treatment algorithm that employs diagnostic tests, it is important to minimize misclassification of animals that may benefit from treatment. This type of misclassification may lead to a rise in the incidence of clinical mastitis, relapses, or the creation of reservoirs of infection.

The dry period is a key time for mastitis prevention. The prevention of new infections, as well as the treatment of existing infections, has been traditionally accomplished by blanket antimicrobial dry cow therapy. Gram positive infections, most commonly,

Staphylococcus aureus andStreptococcus uberis, are common causes of existing intramammary infections at the time of dry-off (16). When cows are infected at the time of dry-off, the objective of the dry cow management program should be to cure the existing infection. When cows are uninfected at dry-off, the goal of dry cow management is prevention of new infections (16). By using the information about the infection status of a cow at the time of dry-off, a rational approach to dry cow therapy can be employed.

In approximately 80% of cases, antibiotic residues in milk can be traced back to mastitis treatments given during lactation or the dry period (17). These concerns have prompted researchers to look at ways to selectively target dry cow therapy to animals that will benefit most from treatment. The prudent use of long-acting dry cow therapy would

6 involve treating only infected cows that are expected to respond to antibiotics and leaving uninfected cows (18) and cows infected with organisms expected to be refractory to antibiotics, without treatment. As uninfected cows are still at risk of developing a new infection during the dry period, producers may choose to use non-antibiotic methods of prevention, namely internal teat sealants (ITS). In 2006, Sanfordet al. did not identify any significant treatment effect when use of an ITS alone was compared with use of cloxacillin alone in cows without evidence of infection late in the lactation period (19).

A selective dry cow therapy program requires the ability to identify cows that are most likely to benefit from treatment with antibiotics at the time of dry-off. The costs of selective dry cow therapy programs were investigated by Huijps and Hogeveen in 2007.

Their model found that appropriate selection of cows that would receive antibiotics at dry- off resulted in a lower average cost of mastitis around the dry period. They also found that the selection of cows was more important than the specific antibiotic used (20). Practically, a diagnostic test used for this purpose should be readily available on the farm, easy to use and interpret and inexpensive. In 2006, Sanfordet al. reported that the California Mastitis Test had a sensitivity of 70% when used to detect mastitis causing pathogens at the time of dry-off

(21). In 2008, a study done by Torreset al. evaluated the use of clinical mastitis history and

SCC from monthly Dairy Herd Improvement (DHI) records for the identification of infected and uninfected cows at dry-off. Various thresholds of cell counts were used to classify cows as infected, with resulting sensitivities calculated between 62.5% and 85.1% (18). Somatic cells are elevated when an inflammatory response occurs in the udder, therefore, they are commonly used to distinguish between infected and uninfected quarters (17). Somatic cell

7 count data are readily available to dairy producers that participate in DHI programs, making it a feasible tool that could be used in selecting cows that would benefit most from dry cow antimicrobial products. Torreset al. concluded that cows with intramammary infections at the time of dry-off could be adequately identified by combining information from somatic cell count and clinical mastitis history but cautioned that decisions regarding selection criteria and adaption of selective dry cow therapy depend on the prevalence of intramammary infections in a herd and the type of microorganisms involved (18).

1.2. Microbiologic Tests for Mastitis

Microbiology is considered the gold standard test for determining the infection status of a cow (21). The Laboratory and Field Handbook on Bovine Mastitis was published by the

National Mastitis Council in 1987 and was revised in 1999 (22,23). The NMC guidelines place culture results in the categories of not significant, questionable significance, probable significance and highly significant. The species isolated, the number of colonies isolated and whether or not the organisms isolated are pure or mixed cultures determine which category results are placed in. The purpose of these publications was to bring uniformity to microbiological procedures used for diagnosing bovine mastitis, and are currently used for diagnostic protocols in many mastitis laboratories. In Canada, laboratories using these standard microbiologic procedures are readily available to many dairy producers that wish to identify causative organisms of mastitis cases. When designing a treatment regimen, having information on the causative organism in order to choose an antimicrobial with an appropriate spectrum of activity is important (24). However, some difficulties may be encountered should a producer wish to make a treatment decision based results obtained by

8 these standard procedures. Time taken to receive a result, distance to or availability of a local laboratory, and cost of cultures are reasons why a producer may choose to treat a cow empirically rather than base a decision on culture results.

Two types of rapid microbiologic tools currently available in Canada are the

University of Minnesota Easy Culture Systemn, developed by the Laboratory of Udder

Health at the University of Minnesota Veterinary Diagnostic Laboratory (25), and the 3M

Petrifilm, developed by 3M Microbiology in St. Paul, Minnesota (26). The University of

Minnesota Easy Culture System plates can be divided in half or thirds. The Bi-plate has one half that contains a proprietary Factor medium which is selective for gram positive bacteria, and one half that contains MacConkey medium for the identification of gram negative bacteria. The Tri-plate contains both Factor and MacConkey media as well as a third modified TKT medium that is selective forStreptococcus species (25). A variety of 3M

Petrifilms exist for identifying various categories of microorganisms. In this study, three types were evaluated, the Aerobic Count Plate (AC), the Coliform Count Plate (CC) and the

Staph Express Plate (STX). The Petrifilm AC plate is a ready-made culture medium that contains Standard Methods nutrients, a cold water gelling agent and an indicator dye that facilitates colony counting and is used for counting aerobic bacteria. The Petrifilm CC plate contains Violet Red Bile nutrients, a gelling agent and an indicator dye that facilitates colony counting. The top film traps gas produced by lactose fermenting coliforms; gas trapped around red colonies indicates confirmed coliforms. The STX plate contains a coldwater- soluble gelling agent, in addition there is a chromogenic, modified Baird-Parker medium in the plate that is selective and differential forS. aureus. Red-violet colonies on the plate are

9 reported asS. aureus. In cases where the color of the colonies is not easily identified or when colonies other than red-violet are present on the plate, 3M has made available a disk that is applied to the plate which identifiesS. aureus. The STX disk contains a DNAse assay that reacts with S. aureus to form pink zones around the colonies (26). All of these tests are capable of rapidly identifying groups of mastitis causing bacteria, and their use on the farm may permit treatment decisions to be made quickly.

It is widely accepted that bacteriologic culture is an imperfect ‘gold standard’ test.

More recent work completed by the Canadian Bovine Mastitis Research Network (CBMRN) has been done to define operating characteristics of various definitions of intramammary infections. A conjoint analysis of expert opinions completed by Andersenet al. (27) with the help of the Mastitis Research Workers’ Conference using sets of three samples taken one week apart yielded 2 consensus definitions for intramammary infections: 1) the organism of interest was isolated on the test day with at least 10 colonies (1000 cfu/mL), and 2) the organism of interest was isolated at least twice in a 3 week period (27). Dohooet al. then expanded on this work by evaluating a set of rules for classifying the infection status of a quarter, based on a single milk sample (28). They found that for all species, exceptS. aureus, the sensitivity of 12 different definitions evaluated was <90% and often <50%. It was concluded that if identifying as many existing infections as possible is important, then the criteria for considering a quarter positive should be a single colony (from a 0.01 mL milk sample) isolated (28).

10 13. Study Objectives

The overall objective of this study was to determine which of the two microbiologic tools under consideration that would yield the greatest diagnostic information in a timely manner with the fewest errors. The first goal was to evaluate two tests, the University of

Minnesota Bi-plate and the 3M Petrifilm Aerobic and Coliform Count Plates as potential rapid on-farm culture systems and to assess their ability to categorize milk samples from clinical mastitis cases into two treatment categories. A treatment category, for all samples positive for gram positive bacterial growth and a no treatment category for samples showing no bacterial growth or growth of only gram negative organisms. Secondly, the two culture systems were evaluated on their ability to provide further diagnostic information on species or species grouping that could potentially aid in management decisions. This was done by comparing microbiological results of the University of Minnesota Tri-plate and the 3M

Petrifilm Staph Express Plate to standard culture techniques for identification of clinical mastitis caused byStaphylococcus aureus and evaluating the Tri-plate’s ability to differentiateStreptococcus spp. from other gram positive organisms. The third goal of the study was to complete a preliminary assessment of the tests for their potential to be used as part of a selective dry cow therapy program. This was accomplished by evaluating the ability of the 3M Petrifilm and the University of Minnesota Tri-plate to classify a cow as either infected or uninfected at the time of dry-off. For this objective, each test was evaluated individually as well as in combination with somatic cell count data obtained from monthly

Dairy Herd Improvement records.

11 1.4. References

1. Erskine, R. J., S. Wagner and F. J. DeGraves. 2003. Mastitis therapy and pharmacology. Vet. Clin. North Am. Food Anim. Pract. 19:109-138.

2. Ruegg P. L. 2003. Investigation of mastitis problems on-farms. Vet. Clin, of North Am. Food Anim. Pract. 19:47-73.

3. Olde Riekerink, R., H. Barkema, D. Kelton and D. Scholl. 2008. Incidence rate of clinical mastitis on Canadian dairy farms. J. Dairy Sci. 91:1366-1377.

4. Leger, D., D. Kelton, K. Lissemore, R. Reid-Smith, W. Martin and N. Anderson. 2003. Antimicrobial drug use by dairy veterinarians and free stall dairy producers in Ontario. Page 318-319 in Natl. Mastitis Counc. Ann. Mtg. Proc., Fort Worth, TX. Natl. Mastitis Counc., Inc., Madison, WI.

5. Health Canada. 2003. Antimicrobial resistance: Keeping it in the box! Health Policy Research Bulletin Issue 6. 2007:40.

6. Smith, K. L., D. A. Todhunter and P. S. Schoenberger. 1985. Environmental mastitis: Cause, prevalence, prevention. J. Dairy Sci. 68:1531-1553.

7. Hogan, J. and K. L. Smith. 2003. Coliform mastitis. Vet. Res. 34:507-519.

8. Neeser, N. L., W. D. Hueston, S. M. Godden and R. F. Bey. 2006. Evaluation of the use of an on-farm system for bacteriologic culture of milk from cows with low- grade mastitis. J. Am. Vet. Med. Assoc. 228:254-260.

9. Keefe, G. P. and K. E. Leslie. 1997. Therapy protocols for environmental streptococcal mastitis. Page 75-86 in Proceedings of a symposium on udder health management for environmentalStreptococci . Guelph, ON.

10. Leslie, K. E., J. T. Jansen and G. H. Lim. 2002. Opportunities and implications for improved on-farm cow-side diagnostics. Page 147-160 in DeLaval International Hygiene Symposium Proceedings, Kansas City, MO.

11. Sears, P. M. and K. K. McCarthy. 2003. Diagnosis of mastitis for therapy decisions. Vet. Clin. North Am. Food Anim. Pract. 19:93-108

12. Canadian Dairy Information Centre. 2008. Dairy facts and figures. 2008:1. Online. Available: http://www.dairyinfo.gc.ca/_english/dffrindex/.

13. Miller, R. H., H. D. Norman and L. L. M. Thornton. 2007. Somatic cell counts of

12 milk from DHI herds during 2007. USDA AIPL Research Report. 2008:6.

14. Godden, S., A. Lago, R. Bey, K. Leslie, P. Ruegg and R. Dingwell. 2007. Use of on-farm culture systems in mastitis control programs. Page 1-9 in Natl. Mastitis Counc. Reg. Mtg. Proc., Visalia, CA. Natl. Mastitis Counc. Inc., Madison, WI.

15. Silva, B. O., D. Z. Caraviello, A. C. Rodrigues and P. L. Ruegg. 2005. Evaluation of Petrifilm for the isolation of staphylococcus aureus from milk samples. J. Dairy Sci. 88:3000-3008.

16. Bradley, A. J. and M. J. Green. The importance of the nonlactating period in the epidemiology of intramammary infection and strategies for prevention. 2004. Vet. Clin. North Am. Food Anim. Pract. 20:547-68.

17. Schukken, Y.H., D. J. Wilson, F. Welcome, L. Garrison-Tikofsky and R. N. Gonzalez. 2003. Monitoring udder health and milk quality using somatic cell counts. Vet. Res. 34(5):579-96.

18. Torres A. H., P. J. Rajala-Schultz, F. J. Degraves and K. H. Hoblet. 2008. Using dairy herd improvement records and clinical mastitis history to identify subclinical mastitis infections at dry-off. J. Dairy Res. 75:240-7.

19. Sanford C.J., G. P. Keefe, I. R. Dohoo, K.E. Leslie, R. T. Dingwell, L. DesCoteaux and H. W. Barkema. 2006. Efficacy of using an internal teat sealer to prevent new intramammary infections in nonlactating dairy cattle. J. Am. Vet. Med. Assoc. 228:1565-73.

20. Huijps K. and H. Hogeveen. Stochastic modeling to determine the economic effects of blanket, selective, and no dry cow therapy. 2007. J. Dairy Sci. 90:1225-34.

21. Sanford C.J., G. P. Keefe, J. Sanchez, R. T. Dingwell, H. W. Barkema, K. E. Leslie and I. R. Dohoo. 2006. Test characteristics from latent-class models of the California Mastitis Test. Prev. Vet. Med. 77:96-108..

22. National Mastitis Council. 1987. Laboratory and Field Handbook on Bovine Mastitis. 1st ed. W. D. Hoard and Sons Co., Fort Atkinson, WI.

23. National Mastitis Council. 1999. Laboratory and Field Handbook on Bovine Mastitis. 2nd ed. Natl. Mastitis Counc. Inc., Madison, WI.

24. Constable, P. D. and D. E. Morin. 2003. Treatment of clinical mastitis: using antimicrobial susceptibility profiles for treatment decisions. Vet. Clin. North Am. Food Anim. Pract. 19:139-155.

13 25. Laboratory for Udder Health. 2000. Minnesota Easy Culture System H Handbook. Minnesota Veterinary Diagnostic Laboratory, University of Minnesota, Saint Paul, MN.

26. 3M Microbiology. 2005. 3M Petrifilm Interpretation Guide. 3M Microbiology, Saint Paul, MN.

27. Andersen S., I.R. Dohoo, R. Olde Reikerink, H. Stryhn and Mastitis Research Workers’ Conference. Diagnosing intramammary infections: Evaluation expert opinions of the definition of intramammary infection using conjoint analysis. 2010. J. Dairy Sci. 93: 2966-2975.

28. Dohoo I.R., J. Smith, S. Andersen, D.F. Kelton, S. Godden, and Mastitis Research Workers’ Conference. Diagnosing intramammary infections: Evaluation of definitions based on a single milk sample. 2011. J. Dairy Sci. 94: 250-261.

14 CHAPTER 2. LABORATORY EVALUATION OF 3M PETRIFILM AND UNIVERSITY OF MINNESOTA BI-PLATES AS POTENTIAL ON-FARM TESTS FOR CLINICAL MASTITIS

J. L. McCarron, G. P. Keefe, S. L. B. McKenna, I. R. Dohoo, and D. E. Poole

Published in Journal of Dairy Science 2009. 92:2297-2305

Revised version June 2012

15 2.1. Abstract

The objective was to determine test characteristics and compare 2 potential on-farm culture systems for clinical mastitis, the Minnesota Easy Culture System II Bi-plate

(University of Minnesota, Minneapolis, MN) and the Petrifilm (3M, Minneapolis, MN). The tests were evaluated using clinically positive mastitic milk samples (n = 282) to determine their ability to differentiate appropriate treatment groups; all cases that had gram positive growth were considered treatment candidates (n = 161) while cases that grew gram negative organisms only or yielded no bacterial growth were classified as no treatment (n = 121). For

Petrifilm, both undiluted and 1:10 diluted milk samples were used. To create the two treatment categories, 2 types of Petrifilms were used, Aerobic Count (AC) and Coliform

Count (CC). Both Bi-plates and Petrifilms were read after 24 h of incubation. Statistical analysis was conducted at various colony count thresholds for the Petrifilm test system. The combination of Petrifilms that had the highest sensitivity classified a case as gram negative if there were > 20 colonies present on the CC. If there were < 20 colonies present on the CC and > 5 colonies present on the AC, a case would be classified as gram positive. The Bi­ plate had a sensitivity of 97.9% and a specificity of 67.5%. The Petrifilm test system had a sensitivity of 93.8% and a specificity of 67.8%. There was no difference in the sensitivities between the tests. When masked readers assessed the plates and Petrifilms, Kappa values were 0.75 for Bi-plates and 0.84 and 0.86 for AC and CC Petrifilms, respectively. The Bi­ plate and Petrifilm were successfully able to categorize clinical cases of mastitis into 2 treatment categories, based on their ability to detect the presence of a gram positive

16 organism. Neither method had the ability to determine if a sample was contaminated i.e. had

3 or more colony types. The results of this study indicate that both tests were able to appropriately categorize cases, which could potentially result in a reduction in the quantity of antibiotics used to treat clinical cases of mastitis.

2.2. Introduction

Mastitis remains the most costly infectious disease to the dairy industry and is the

most frequent cause of antibacterial use on dairy farms (1). A recent comprehensive study of

clinical mastitis on Canadian dairy farms found the mean incidence rate of clinical mastitis at

23.0 cases per 100 cow-years (2).

In Canada, mastitis treatment accounts for more than half of all antibiotics used by

dairy producers (3). Much of the antibiotic used to treat clinical mastitis may not be justified

because of high self cure rates and low efficacy of treatments against certain pathogens. For

example, the use of intramammary antimicrobials for the treatment ofEscherichia coli

clinical cases was most often ineffective due to the short duration of infections and high

spontaneous cure rates (4,5). There are public concerns that the overuse of antibiotics in

agriculture may lead to antimicrobial resistance in humans (6). For these reasons, the judicious use of antibiotics by veterinarians and producers continues to be emphasized

throughout the dairy industry.

Clinical mastitis is caused by a wide range of bacteria. When designing a treatment

regimen, having information on the causative organism in order to choose an antimicrobial

with an appropriate spectrum of activity is important (7). Still, when a dairy producer

encounters a clinical case of mastitis, many times, a diagnosis of the causative organism is

17 not obtained prior to treatment. One reason for not obtaining this information is the delay that exists between sampling an affected quarter and receiving results from a diagnostic laboratory. If producers experience delays in receiving results, they may choose to treat empirically rather than wait for a culture result (8).

A recent Canadian study reported that 43.9% of 3,033 milk samples submitted from cows with clinical mastitis yielded no bacterial growth (2). These negative results may be due to spontaneous clearance of pathogens, cyclical sheddingS. aureus,of phagocytosis of bacteria within the milk sample, and the presence of too few bacterial colonies to yield a diagnosis (9). In these situations, the treatment with an antimicrobial may not be warranted.

Additionally, even if a milk sample is positive for a pathogen, the characteristics of the pathogen must be considered before making a treatment decision (7).

There is a need for a more targeted approach to clinical mastitis therapy (10). One method to target therapy is to use an on-farm culture system. By using an on-farm system, producers would be able to make informed mastitis treatment decisions that may lead to the reduction in the use of antimicrobials. On-farm tests that are rapid, sensitive and have high negative predictive values would be valuable tools for producers. Evidence that on-farm testing could be used to make clinical mastitis therapy decisions was described by Leslieet al. in 2002 (11). Selective growth media may be useful on-farm tools for determining the major pathogen categories that cause mastitis, thus enabling a producer to choose an appropriate treatment. Many dairy farms in the US are large enough to set up and maintain milk culturing laboratories. The results generated from these on-farm laboratories could be used to develop treatment protocols and monitor long-term mastitis situations (12). Godden

18 et al. reiterated the need for the adoption of rapid on-farm culture systems that would allow producers to make informed treatment decisions (13). By treating fewer cases with antibiotics and discarding less milk that contains residues, the cost per case of mastitis was reduced (14). Tests that will play a part in reducing the economic impacts of mastitis, as well as promote the prudent use of antimicrobials on dairy farms, need to be validated (13).

There has been a perceived increase in the adoption of on-farm culture systems for the selective treatment of clinical mastitis on dairy farms in the US (15). In Canada, the demographics of the dairy industry are different than in the US. The average dairy herd size in Canada in 2007 was 68 cows (16), whereas, the average herd size in the US in 2007 was

125 cows (17). The size of a herd may play a role in the uptake and type of on-farm test system that is suitable.

The Minnesota Easy Culture System II Bi-plate and the 3M Petrifilm both have the capability of distinguishing samples with no bacterial growth, gram positive infections and gram negative infections. The aim of this study was to evaluate, in a laboratory setting, the ability of the tests to categorize milk samples from clinical mastitis cases into potential treatment categories.

2 3 . Material and methods

2.3.1. Samples

A convenience sample of 21 farms located close to the Atlantic Veterinary College was recruited. Milk samples (n = 282) were collected by trained dairy producers from cows with clinical mastitis. Clinical mastitis was defined as milk that appeared abnormal with or without other local or systemic signs. Samples were collected after routine teat preparation

19 using the technique described by NMC (18). Samples were refrigerated on the farm, never frozen, and picked up daily by technicians. The total time elapsed between farmer sample collection and laboratory culture did not exceed 36 h and was typically < 24 h.

2.3.2. Gold Standard

Gold standard bacteriological cultures were performed according to the Laboratory

Handbook on Bovine Mastitis (18). Samples were classified as having significant growth if the growth was considered of ‘probable significance’ or ‘highly significant’ based on

National Mastitis Council Guidelines for significance (19). Disposable plastic loops were used to streak 10 pL of each sample onto blood agar and MacConkey plates. Plates were incubated at 35 °C for 24 h. The standard laboratory plates were read by a milk laboratory technician, and additional tests were conducted to confirm the genus and species of the organism present in each case. Results of these cultures were then used to assign cases to 2 treatment categories. The treatment category consisted of all samples that had significant growth of a gram positive organism only, while the no treatment category consisted of all samples that resulted in no significant microbial growth or growth of a gram negative organism only. Samples that grew a yeast or mold (no bacterial growth) were included in the no treatment category. Samples that had 2 colony types were considered mixed growth and samples with 3 or more colony types were considered contaminated. Mixed growth and contaminated samples were not used in the decision models; however, the impact of these samples was evaluated.

2.3.3.3M Petrifilm

20 The second media system used was the 3M Petrifilm Aerobic Count (AC) and the 3M

Petrifilm Coliform Count (CC). The Petrifilm AC plate is a ready made culture medium that contains Standard Methods nutrients, a cold water gelling agent and an indicator dye that facilitates colony counting (20) and is used for counting aerobic bacteria. The Petrifilm CC plate contains Violet Red Bile nutrients, a gelling agent and an indicator dye that facilitates colony counting. The top film traps gas produced by lactose fermenting coliforms; gas trapped around red colonies indicates confirmed coliforms (20). A 1 mL aliquot of each sample was plated on the Petrifilm AC and CC plates. In addition, milk samples were diluted

1:10 with sterile water and plated on 2 more Petrifilm AC and CC plates. All plates were i incubated at 35 °C for 24 h. Petrifilms were used in combination to create treatment categories of no growth, gram positive growth only and gram negative growth only. Each

Petrifilm was read by the technician and categorized as positive if there were 20 or more colonies present. Colony growth on both the AC and CC Petrifilms was classified as gram negative and assigned to the no treatment category. Colony growth on only the AC Petrifilm was classified as a gram positive and assigned to the treatment category. If there were < 20 colonies on both Petrifilms the sample was categorized at no treatment.

2.3.4. Minnesota Easy Culture System II Bi-plate

The Minnesota Easy Culture System II Bi-plate, developed by the University of

Minnesota Laboratory for Udder Health, is a culture plate that is divided in half. One half contains a proprietary Factor medium that is selective for gram positive bacteria, and the other half contains MacConkey medium for the identification of gram negative bacteria (21).

The media were inoculated according to the manufacturer’s recommendations. Sterile cotton

2 1 tipped swabs were saturated in milk and used to swab one-half of the plate, then re-dipped in the sample to swab the other half. Plates were incubated in a 35 °C incubator for 24 h before being read. Depending on which side of the Bi-plate was positive for growth, the Bi-plates were assigned to a treatment category. Growth (1 or more colonies) on the Factor medium only was considered gram positive growth and assigned to the treatment category. Growth (1 or more colonies) on the MacConkey medium only, was considered gram negative and assigned to the no treatment category. If there was growth on both media of the Bi-plate, the plate was considered contaminated, but due to the presence of gram positive organisms these plates were assigned to the treatment category. Finally, if there was no growth on either media, the sample was assigned to the no treatment category.

2.3.5. Inter-reader Agreement

All Petrifilms were read by a milk laboratory technician in addition to 3 ‘masked’ readers with limited microbiology training to determine the inter-reader agreement beyond chance (kappa). Readers were asked to identify the presence of colony growth and if positive, to record the colony count as greater or fewer than 20 colonies. Bi-plates were read by the same laboratory technologist and 4 ‘masked’ readers. All readers were asked to identify the presence of bacterial colonies (1 or more) and which side of the plate the growth was on. Inter-reader agreement was assessed by calculating pooled kappa statistics for both test systems.

2.3.6. Statistical Analysis

All results were analyzed using Intercooled Stata 9 (22). The Petrifilm test system

(AC and CC combined) was first evaluated to determine the performance of the test on

2 2 undiluted versus diluted milk samples. This was done by comparing the test characteristics and associated 95% confidence intervals of each test system. Using the results obtained from the diluted milk samples, the Petrifilm treatment decision results were compared to those of the gold standard to determine the sensitivity (Se), specificity (Sp), positive predictive value

(PPV) and negative predictive value (NPV) of the Petrifilm test system.

Further evaluation of the Petrifilms was carried out to determine the effect that different colony count threshold values would have on the Petrifilm test systems’ sensitivity and specificity. This was done using the results from only the diluted sample Petrifilms. To evaluate the effects of different cut off values, the technologists’ results were first compared to those of the automated 3M Petrifilm reader to determine the agreement between the 2 reading methods. Since the level of agreement for each type of Petrifilm was very high, the results of the automated reader were used to determine test sensitivity and specificity at various colony-forming-unit cut points.

Data obtained by the technologist and the automated reader were assessed using

McNemar’s test for bias before calculating kappa statistics for both the AC and CC

Petrifilms. Since the level of agreement for each type of Petrifilm was very high, the results of the automated reader were used to create sensitivity vs. specificity plots (also called 2 graph receiver operating characteristics) to evaluate test sensitivity and specificity at various colony-forming-unit cut-points.

For assessment of the Bi-plate system, only results recorded by the laboratory technician were used. Sensitivity, Sp, PPV and NPV were calculated by comparing the gold

23 standard treatment and no treatment classifications to the Bi-plate treatment and no treatment classifications.

2.4. Results

2.4.1 Samples

A total number of 282 fresh milk samples were received. Table 2.1 contains a summary of the results of gold standard cultures. There were 280 samples that had complete

Bi-plate records (gold standard culture and read by the laboratory technician). The prevalence of treatment candidates was 54.2% in the samples that were used to calculate test characteristics for the Bi-plates. There were 275 samples with complete records for undiluted milk samples cultured on Petrifilms.

After 6 weeks of sample collection, readers noted that 1 mL of undiluted milk, plated on the Petrifilms, resulted in some plates that were difficult to interpret by the technician; milk clots and cases of heavy bacterial growth resulted in difficulty identifying individual colonies. At that point, after 55 samples had been plated already, using the same milk samples, a second series of Petrifilms with a 1 in 10 dilution (equivalent to 100 uL of milk and 900 uL of sterile diluent on the film) was initiated, in addition to the undiluted Petrifilms and Bi-plates. Therefore, the total number of diluted samples cultured on Petrifilms and read by the technician was not 275 but 220. The prevalence of treatment candidates was 56.5% in the samples that were used to calculate test characteristics for the diluted Petrifilms.

2.4.2 3M Petrifilm Undiluted Versus Diluted Samples

In total, there were 275 undiluted clinical milk samples that were cultured on the

Petrifilm system. Seventeen samples were removed from the analysis because they were

24 diagnosed as mixed growth or contaminated on the gold standard culture. Of these 17 samples, 12 were characterized as gram positive growth using the Petrifilms and would have been treated if Petrifilm results were to be used for treatment decisions. There were an additional 4 samples that did not have complete records that were not included in this analysis.

There were 220 samples read by the technician that were diluted and plated on both the AC and CC Petrifilms. Of these, 17 were removed from the interpretation due to mixed growth or contamination of the gold standard plates. The others had incomplete readings by the automated reader. If these samples had been included, 11 were classified as gram positive based on the Petrifilms and placed in the treatment category, leaving 200 samples for comparison with culture results.

When the Petrifilms were used with diluted milk samples, the sensitivity and negative predictive values were not different from undiluted samples. There were differences in specificity (72.6% versus 92.9%) and positive predictive values (77.7% versus 93.8%), both of which favored using the diluted milk samples. For this reason, further statistical analysis to choose appropriate colony count thresholds for the Petrifilms was done using only the diluted milk samples.

2.4.3 Agreement between Manual and Automated Reading of Petrifilms

Table 2.2 illustrates the agreement between the technologist and automated Petrifilm reader for AC and CC Petrifilms using diluted milk samples. Before calculating the level of agreement between the 3M automated reader and the technician, McNemar’s tests for significance were performed. The exact McNemar significance probabilities were 0.13 and

25 1.0 for the AC and CC Petrifilms, respectively. The non significant test results indicate that the two proportions positive did not differ, therefore kappa could be evaluated. The calculated agreement between the technician and the automated reader was 96.7% with a kappa value of 0.93 for the AC Petrifilm. For CC Petrifilm, the calculated agreement was

98.6% with a kappa value of 0.92. Because these levels of agreement were high, the colony count values obtained from the automated reader were used to evaluate the effect that various thresholds would have on test characteristics.

2.4.4 Evaluation of Colony Count Thresholds

As with the previous analysis, all samples that were contaminated on the gold standard were removed from the data set, leaving 200 AC and CC Petrifilms with complete records from the automated reader. Sensitivity versus specificity plots were created for various combinations of colony count thresholds ranging from 1 to 30, in increments of 5.

Figure 2.1 contains a truncated graph of the test characteristics (Se and Sp) for each combination of colony count thresholds.

The combination of Petrifilms that resulted in the optimum sensitivity used a threshold of <20 colonies on the CC Petrifilm and >5 colonies on the AC Petrifilm to consider the test positive for gram positive growth (Figure 2.1). Values above 20 (25 and

30) for the CC were evaluated, but there was virtually no gain in sensitivity using a threshold of 25 and only a slight gain at 30 (data not shown). At every level of CC threshold evaluated, the greatest decrease in the Petrifilm test system’s sensitivity was observed when thresholds >5 were used on the AC. The test characteristics of the Petrifilm system that uses a threshold of <20 colonies on the CC plate and >5 colonies on the AC plate are shown in

26 Table 2.3. Samples in which the CC plate contained 20 or more colonies would be considered gram negative and be interpreted as negative (no treatment). Samples in which the AC plate has fewer than 5 colonies would be considered no growth and also interpreted as negative (no treatment). All other scenarios would be classified as positive (treatment).

There were only 7 samples out of the 200 that were diagnosed as needing treatment on the gold standard culture that the Petrifilm test system diagnosed as no treatment needed.

Six of the 7 samples grew < 5 colonies on the AC Petrifilm and 1 gram positive sample was misclassified as a coliform. Twenty-six samples were classified as needing treatment based on Petrifilms, while according to the gold standard, they should not have been treated. All of the cases that fit this description yielded no bacterial growth on the gold standard, except for

4. Of the 4,1 was positive for Serratia and the remaining 3 were positive forP. multocida.

These 2 gram negative pathogens were not coliforms and would not be identified by the CC

Petrifilm.

2.4.5 Minnesota Easy Culture System'll Bi-Plate

Eighteen of the 280 samples that were cultured on the gold standard exhibited mixed growth or contamination and were not used in the evaluation of the Bi-plates, leaving 262 samples for comparison with culture results. Seventeen of the 18 samples would have been classified as needing treatment based on the Bi-plate results. Table 2.4 illustrates the test characteristics of the Bi-plate when read by the laboratory technician who used a threshold of

1 colony to consider the plate positive.

The Bi-plate was highly sensitive because there were only 3 samples that should have been included in the treatment category, based on culture results, that the Bi-plate classified

27 as no treatment. The gold standard results of these 3 samples were: 1 coagulase negative

Staphylococcus spp. (CNS), 1 S. uberis and 1S. dysgalactiae. The reduced specificity was the result of 39 samples being classified as needing treatment by the Bi-plate that did not require treatment according to the gold standard. Most (28) of these samples had no bacterial growth on the gold standard, the remaining samples includedE. coli, 7 2 P. multocida, and 2 yeast.

2.4.6 Inter-reader agreement

Agreement between readers was assessed by calculating kappa values for each test system. All readers were masked regarding the results from the gold standard and the other participants. For Bi-plates, the kappa value was calculated based on the ability to diagnose no growth, gram positive growth only, gram negative growth only or contamination (growth on both sides of the Bi-plate). The Bi-plates were read by 5 individuals and the combined kappa value was 0.76. Only 4 of the 5 individuals evaluated all of the Petrifilms. The kappa value for the AC Petrifilms using diluted milk samples was 0.84, based on the ability to detect colony growth with 5 or more colonies present. The kappa value for the CC Petrifilms was 0.86, based on their ability to identify a CC Petrifilm with 20 or more colonies present.

2.5. Discussion

The most important part of an effective mastitis control program is the use of prevention strategies; however, appropriate therapy of individual clinical cases remains important (1). Dairy producers frequently experience cases where the empirical treatment of mastitis with intramammary products is ineffective, due in part to inappropriate therapeutic choices. Fifty-eight percent of submitted clinical mastitis cases were culture negative or

28 gram negative pathogens (2). Others reported that pathogen profiles did not justify the use of labeled intramammary antibiotics for 50% to 80% of clinical mastitis cases (23).

Having a diagnostic tool suitable for on-farm use would eliminate the empirical treatment of clinical cases and allow producers to make informed decisions, thereby decreasing the amount of antibiotics used unnecessarily. This study performed a thorough laboratory evaluation comparing available technologies, which is the first step that is required before such tools can be recommended in the field. Specifically, the abilities of the

University of Minnesota Bi-plate and the 3M Petrifilm to correctly place milk samples into appropriate treatment categories were evaluated.

In this study, the 282 milk samples received were considered independent. However, the samples originated from 21 farms and in some cases there were multiple samples submitted from the same cow. The analysis presented did not take into consideration the hierarchical structure of the data, thus the effect of clustering was ignored. One common effect of ignoring clustering is that the standard errors of parameter estimates (in this study

Se, Sp, NPV and PPV) may be underestimated (24).

The distribution of culture results for samples used was similar to that of a recent study of the incidence rate of clinical mastitis on Canadian dairy farms (2). Here, 30.7% of the samples cultured resulted in no bacterial growth, whereas in the nation-wide study, 43.9% of samples submitted were culture-negative.S. aureus was the most frequently isolated pathogen isolated in both studies, but comprised a higher percent of the samples (16.8%) in our study compared to the national study (10.3%). Gram negative pathogens(E. coli,

Pasteurella, Klebsiella, Serratia andEnterobacter spp.) accounted for 12.4% of the samples

29 received in our study, whereas, 13.9% of the samples in the Canada-wide study were positive for gram negative pathogens (2).

The Petrifilm system combined both the AC and CC plates to determine if a sample was positive for gram positive growth and a candidate for treatment. A previous trial conducted by Silvaet al. in 2004 used the Petrifilm culture system on a 600 cow dairy that was experiencing a herd problem with clinical mastitis, excessive use of antibiotics, and extended days of milk discard (14). When using the Petrifilm on-farm culturing as part of a treatment protocol, the farm was able to decrease days that milk was kept out of the tank, number of mastitis tubes used per case, and the percent of cases receiving intramammary antibiotics.

There were no differences in the Se or NPV between diluted and undiluted milk samples on the Petrifilm system, even though the undiluted system used 10 times the volume of milk. Dilution of the milk samples resulted in higher Sp and PPV. The increased Sp and

PPV were the result of less frequent misclassification of gold standard negative samples as positive on Petrifilm when diluted samples were used. The potential explanation of this difference is that the individuals looking at the undiluted Petrifilms were confusing debris from the whole clinical milk samples with actual colony forming units, resulting in a false positive diagnosis. For this reason, a more in depth analysis inta colony-forming-unit thresholds was done using only the results of the diluted milk samples.

There were 26 tests (13% of samples) that the Petrifilm classified as treatment positive that were diagnosed as no treatment by the gold standard. This can be explained by the increased volume of milk plated on each Petrifilm, - 100 pL of milk on the Petrifilm as

30 opposed to 10 pL on the gold standard. There are gram negative pathogens that cause mastitis that are not coliform bacteria. These organisms will always be classified as treatment candidates using the Petrifilm system in this manner.

For application in the field, the usefulness of this test system can be best evaluated by looking at its predictive values. Of particular interest to a dairy producer is the NPV of a test.

This is the probability that given a negative test, the cow does not need to be treated. The clinical consequences of not treating an animal that should have been treated outweigh the consequences of treating a case that did not require antimicrobials. In a scenario where an animal with clinical mastitis tests negative and the decision is made not to treat the cow, a producer wants to be confident that the correct decision has been made. In the set of 200 diluted milk samples that were used in this study, the NPV was 89.4% and the PPV was

79.1%. These numbers were influenced by the prevalence (56.5%) of treatment positive cases in the group of samples that were diluted and cultured on the Petrifilm.

Predictive values of a test change with different populations of animals tested because they are driven by the true prevalence of disease in the study population as well as the test characteristics, Se and Sp (24). A more accurate estimate of predictive values can be obtained by using proportions of disease positive and disease negative cases from a larger more representative study population. When using the calculated Se and Sp from our study and the proportions of disease positive (p(D+)) for gram positive and proportion of disease negative

(p(E>-» for gram negative or no growth, from Olde Riekerinket al. (2), predictive values were calculated using the formulae:

31 PPV =------^ ------p(D+) *Se+ p(D -) * (1 - Sp)

N P V - P(D~)*Sp p(D-) *Sp + p(D+) * (1 - Se)

The PPV calculated for Petrifilm was 66.1% and the NPV was 90%. This high NPV indicated that a producer can be very confident in a negative test result, as the probability that a negative culture actually needed treatment was 10%. The lower PPV indicated that using the Petrifilm will continue to result in the unnecessary treatment of cases that do not require antimicrobials. Nonetheless, a reduction in the overall percent of cases that are treated can still be realized. Using the Petrifilm system in this set of samples would have resulted in

66.0% of cases being treated (tested positive for gram-positive bacteria using the Petrifilm system), and therefore a potential 34.0% reduction in the amount of antimicrobials used inappropriately i.e. if a producer were to treat only test positive cows as opposed to all cows.

The Petrifilm system was easy to use, but required some training; colony growth was easy for the individuals to assess, especially when diluted samples were used. The agreement between readers (kappa) was used to evaluate this ease of interpretation for both the AC and

CC Petrifilms. The high combined kappa statistic indicated that agreement between individuals was very good.

The University of Minnesota Bi-plate is highly sensitive in its ability to determine treatment categories for clinical cases of mastitis. Previous laboratory work on the validation of the Bi-plate determined a sensitivity of 83.2%, specificity at 74.5%, PPV of 77.4%, and the NPV at 80.9%, using 101 frozen-thawed milk samples from clinical cases (25). A field

32 trial using the Bi-plate determined sensitivity of 96.8% and specificity at 81.8% (13) based on 101 fresh quarter milk samples that were cultured on-farm.

In this study, there were 39 (14%) samples that the Bi-plate classified as treatment candidates that were diagnosed as no treatment by the gold standard. Of these samples, 28 were negative for growth on the gold standard, yet grew at least 1 colony on the Factor medium of the Bi-plates. This could be due to the difference in the volume of milk plated on the standard blood agar plates (10 pL per half plate) versus a saturated cotton swab of milk per half plate on the Bi-plate. The other 9 samples that the gold standard diagnosed as “do not treat” were all gram negative organisms on the gold standard cultureE. (7coli and 2

Pasteurella spp). The 2 Pasteurella spp that were diagnosed by the gold standard were not expected to be classified as gram negative organisms on the Bi-plate as they do not grow on

MacConkey medium. All of theE. coli samples had growth on both sides of the Bi-plate; therefore, due to the presence of a gram positive they were placed in the treatment category.

Withholding treatment may represent a paradigm shift for many dairy producers. As a result, a strategy that minimized under treatment was adopted. Consequently, all plates with growth on both sides of the Bi-plate were considered treatment cases.

In this study, the PPV of the Bi-plate was 78.1%. Again, this was influenced by the prevalence (54.2%) of treatment positive cases in the group of samples that were cultured on the Bi-plate. As in the evaluation of the Petrifilms, our calculated Se and Sp were applied to the larger more representative population that had a slightly lower prevalence of gram positive cases. The resulting PPV was 60.2% and the NPV was 98.5%. Based on this NPV,

33 the probability that a cow with a negative Bi-plate culture actually needed treatment was only

1.5%.

In total, the University of Minnesota Bi-plate diagnosed 63.8% of the samples received as gram positive organisms. Using a Bi-plate culture to determine which cases of clinical mastitis receive antimicrobial therapy could potentially result in a 36.2% reduction in treatment if the current practice of a producer was to treat all clinical cases. This reduction in treatment assumes that all gram positive cases should be treated. There will continue to be unnecessary treatment of some cases that are caused by bacterial species that are not susceptible to intramammary therapy. Still, when compared to the practice of treating all cows with abnormal milk, the use of Bi-plates combined with treatment protocols based on their outcomes has the potential to reduce the amount of antibiotics used on the farm.

The Bi-plate was easy to use and with some training, colony growth was easy for the individuals to interpret. The agreement between readers was used to evaluate this ease of interpretation and the high combined kappa statistic indicated that agreement between the readers was very good.

When comparing these 2 test systems, the Bi-plate seemed to slightly outperform the

Petrifilm in sensitivity and NPV (no statistically significant differences), while there were no numerical differences in specificity and PPV. Neither method had the ability to reliably determine if a sample was contaminated. Should this technology be used on-farm, duplicate samples should be submitted to a diagnostic laboratory for quality control purposes. The Bi­ plate is able to show mixed growth and contamination if there is a combination of gram positive and gram negative organisms present. However, samples that consist of mixed gram

34 positive organisms can not be identified. Therefore, proper training in collection and handling of samples is an important aspect of on-farm culture. Both systems provided accurate results within 24 h of culturing. The impact of delaying treatment for 24 h has not been well-studied, and has the potential to result in reduced treatment outcomes (8). An on- farm evaluation is required to evaluate this statement, as well as the overall impact that on- farm culture may have on cow health, recurrence rates of mastitis, and economic feasibility.

2.6. Conclusion

The University of Minnesota Bi-plate and the 3M Petrifilm were successfully able to categorize clinical cases of mastitis into 2 treatment categories based on their ability to detect the presence of a gram positive organism. Neither test had the ability to reliably determine if a sample was contaminated. Both the Bi-plate and 3M Petrifilm system were highly sensitive tests, 97.9% and 93.8%, respectively. The very high negative predictive values

(96.4% for Bi-plate and 89.4% for Petrifilm) were important attributes of each of these tests which will minimize the number of cases requiring therapy that go untreated. The results indicate that both tests have the potential to appropriately categorize cases, which could result in a reduction in the amount of antibiotics used to treat clinical cases of mastitis.

2.7. Acknowledgements

The authors would like to thank Theresa Andrews, Lloyd Dalziel, Katie Macintosh,

Michael Trenholm, Cynthia Mitchell, Shana Richard and Matt Saab for their technical help with this project. This research was financed by NSERC, Alberta Milk, Dairy Farmers of

New Brunswick, Nova Scotia, Ontario and Prince Edward Island, Novalait Inc., Dairy

Farmers of Canada, Canadian Dairy Network, AAFC, PHAC, Technology PEI Inc.,

35 University de Montreal and University of Prince Edward Island through the Canadian Bovine

Mastitis Research Network.

2.8. Sources and manufacturers

a. 3M Petrifilm. 3M Microbiology. St. Paul, Minnesota. USA.

b. University of Minnesota Easy Culture SystemEL Laboratory for Udder Health. University of Minnesota, St. Paul, Minnesota. USA.

2.9. References

1. Erskine, R. J., S. Wagner and F. J. DeGraves. 2003. Mastitis therapy and pharmacology. Vet. Clin. North Am. Food Anim. Pract. 19:109-138.

2. Olde Riekerink, R., H. Barkema, D. Kelton and D. Scholl. 2008. Incidence rate of clinical mastitis on Canadian dairy farms. J. Dairy Sci. 91:1366-1377.

3. Leger, D., D. Kelton, K. Lissemore, R. Reid-Smith, W. Martin and N. Anderson. 2003. Antimicrobial drug use by dairy veterinarians and free stall dairy producers in Ontario. Page 318-319 in Natl. Mastitis Counc. Ann. Mtg. Proc., Fort Worth, TX. Natl. Mastitis Counc., Inc., Madison, WI.

4. Smith, K. L., D. A. Todhunter and P. S. Schoenberger. 1985. Environmental mastitis: Cause, prevalence, prevention. J. Dairy Sci. 68:1531-1553.

5. Hogan, J. and K. L. Smith. 2003. Coliform mastitis. Vet. Res. 34:507-519.

6. Health Canada. 2003. Antimicrobial resistance: Keeping it in the box! Health Policy Research Bulletin Issue 6. 2007:40.

7. Constable, P. D. and D. E. Morin. 2003. Treatment of clinical mastitis: using antimicrobial susceptibility profiles for treatment decisions. Vet. Clin. North Am. Food Anim. Pract. 19:139-155.

8. Neeser, N. L., W. D. Hueston, S. M. Godden and R. F. Bey. 2006. Evaluation of the use of an on-farm system for bacteriologic culture of milk from cows with low-grade mastitis. J. Am. Vet. Med. Assoc. 228:254-260.

9. Silva, B. O., D. Z. Caraviello, A. C. Rodrigues and P. L. Ruegg. 2005. Evaluation of Petrifilm for the isolation of staphylococcus aureus from milk samples. J. Dairy Sci. 88:3000-3008.

36 10. Keefe, G. P. and E. Leslie K. 1997. Therapy protocols for environmental streptococcal mastitis. Page 75-86 in Proceedings of a symposium on udder health management for environmental streptococci, Guelph, ON.

11. Leslie, K. E., J. T. Jansen and G. H. Lim. 2002. Opportunities and implications for improved on-farm cow-side diagnostics. Page 147-160 in DeLaval International Hygiene Symposium Proceedings, Kansas City, MO.

12. Sears, P. M. and K. K. McCarthy. 2003. Diagnosis of mastitis for therapy decisions. Vet. Clin. North Am. Food Anim. Pract. 19:93-108.

13. Godden, S., A. Lago, R. Bey, K. Leslie, P. Ruegg and R. Dingwell. 2007. Use of on- farm culture systems in mastitis control programs. Page 1-9 in Natl. Mastitis Counc. Reg. Mtg. Proc., Visalia, CA. Natl. Mastitis Counc. Inc., Madison, WI.

14. Silva, B., D. Caraviello, A. Rodrigues and P. Ruegg. 2004. Use of petrifilm for mastitis diagnosis and treatment protocols. Page 52 in Natl. Mastitis Counc. Ann. Mtg. Proc., Charlotte, NC. Natl. Mastitis Counc. Inc., Madison, WI.

15. Lago, A., S. Godden, R. Bey, K. Leslie, R. Dingwell and P. Ruegg. 2006. Validation of the Minnesota easy culture system II: Results from on-farm bi-plate culture versus standard laboratory culture. Page 250-251 in 39th Annu. Proc. Am. Assoc. Bovine Pract., Saint Paul, MN. Am. Assoc. Bovine Pract., Auburn, AL.

16. Canadian Dairy Information Centre. 2008. Dairy facts and figures. 2008:1. Online. Available: http://www.dairyinfo.gc.ca/_english/dffiindex/.

17. Miller, R. H., H. D. Norman and L. L. M. Thornton. 2007. Somatic cell counts of milk from DHI herds during 2007. USDA AIPL Research Report. 2008:6.

18. National Mastitis Council. 1999. Laboratory and Field Handbook on Bovine Mastitis. 2nd ed. Natl. Mastitis Counc. Inc., Madison, WI.

19. National Mastitis Council. 1987. Laboratory and Field Handbook on Bovine Mastitis. 1st ed. Natl. Mastitis Counc. Inc., Arlington, VA.

20. 3M Microbiology. 2005. 3M Petrifilm Interpretation Guide. 3M Microbiology, Saint Paul, MN.

21. Laboratory for Udder Health. 2000. Minnesota Easy Culture System H Handbook. Minnesota Veterinary Diagnostic Laboratory, University of Minnesota, Saint Paul, MN.

37 22. StataCorp. 2005. Intercooled Stata 9.0. College Station, Texas, USA.

23. Roberson, J. R. 2003. Establishing treatment protocols for clinical mastitis. Vet. Clin. North Am. Food Anim. Pract. 19:223-234.

24. Dohoo, I., W. Martin and H. Stryhn. 2009. Veterinary Epidemiologic Research. 2nd ed. VER Inc., Charlottetown, PE, Canada.

25. Hochhalter, J., S. Godden, R. Bey, A. Lago and M. Jones. 2006. Validation of the Minnesota Easy Culture System H: Results from in-lab bi-plate culture versus standard laboratory culture, and bi-plate inter-reader agreement. Page 298 in 39th Annu. Proc. Am. Assoc. Bovine Pract., Saint Paul, MN. Am. Assoc. Bovine Pract., Auburn, AL.

38 Table 2.1.Gold standard microbiology results of samples submitted to be evaluated by Bi­ plates and Petrifilms

Gold Standard Result Samples % Undiluted % Diluted % submitted Samples Samples for Bi- submitted submitted Plate for for Evaluation Petrifilm Petrifilm i Evaluation Evaluation (n = 280) 2 (n = 275) 3 (n = 220) No Bacterial Growth 86 30.7 86 31.3 61 27.7 Staphylococcus aureus 47 16.8 47 17.1 38 17.3 Streptococcus uberis 42 15.0 41 14.9 36 16.4 Escherichia coli 27 9.6 26 9.5 20 9.1 Streptococcus spp. 19 6.8 18 6.5 13 5.9 Streptococcus dysgalactiae 18 6.4 17 6.2 13 5.9 Contaminated or mixed growth 18 6.4 17 6.2 19 8.6 Staphylococcus spp. 10 3.6 10 3.6 9 4.1 Pasteurella multocida 3 1.1 3 1.1 3 1.4 Staphylococcus hyicus 3 1.1 3 1.1 3 1.4 Klebsiella spp. 2 0.7 2 0.7 2 0.9 Streptococcus bovis 1 0.4 1 0.4 0 0 Bacillus spp. 1 0.4 1 0.4 0 0 Corynebacterium spp. 1 0.4 1 0.4 1 0.5 Serratia spp. 1 0.4 1 0.4 1 0.5 Enterobacter spp. 1 0.4 1 0.4 1 0.5 Total 280 100 275 100 220 100 Gold standard results of the 280 samples that were offered to the Bi-plate for evaluation 2 Gold standard results of the 275 undiluted samples that were offered to the Petrifilm for evaluation 3 Gold standard results of the 220 diluted samples that were offered to the Petrifilm for evaluation

39 Table 2.2. Agreement between technician and automated Petrifilm reader for Aerobic Count and Coliform Count Petrifilms using diluted milk samples

Technician AC1 Petrifilm Technician CC2 Petrifilm Automated Plate Negative Positive Total Negative Positive Total Reader < 20 colonies > 20 colonies < 20 colonies > 20 colonies Negative 75 1 76 188 2 190 < 20 colonies Positive 6 129 135 1 20 21 > 20 colonies Total 81 130 211 189 22 211 AC = Aerobic Count 2CC = Coliform Count

40 Table 2.3. Test characteristics of the Petrifilm test system that uses a threshold of <20 colonies on the Coliform Count plate and >5 colonies on the Aerobic Count plate based on its ability to correctly classify a sample as gram positive (n=200 samples)

Test Characteristic (95% Confidence Interval) Sensitivity 93.8% (87.7 - 97.5) Specificity 67.8% (56.9-77.4) Positive Predictive Value 79.1% (71.2-85.6) Negative Predictive Value 89.4% (79.4-95.6)

41 Table 2.4.Test characteristics of the University of Minnesota Bi-plate based on its ability to correctly classify samples as gram positive (n=262 samples)

Test Characteristic (95% Confidence Interval) Sensitivity 97.9% (94.0 - 99.6) Specificity 67.5% (58.3 - 75.8) Positive Predictive Value 78.1% (71.3-83.9) Negative Predictive Value 96.4% (89.9 - 99.3)

42 1 0 0

a

I II ,11 III AC 5 I AC 10 AC 20 AC 5 AC 10 AC 20 AC 10 AC 20

CC 5 CC 10 CC 20 Coni>inat»n of Colony Count Cut-Offe

■ Sensitivity Ei Specificity Figure 2.1. Comparison of sensitivity and specificity for various combinations of colony count thresholds (grouped as 5, 10 and 20 colony forming units respectively) for the diluted Aerobic Count (AC) and Coliform Count (CC) Petrifilms For example, AC5 CC10 means that a sample would be considered negative if there were fewer than 5 colonies on the AC plate and/or 10 or more on the CC plate and positive if there were 5 or more colonies on the AC plate and less than 10 on the CC plate.

43 CHAPTER 3. EVALUATION OF THE UNIVERSITY OF MINNESOTA TRI-PLATE

AND 3M PETRIFILM FOR THE ISOLATION OF STAPHYLOCOCCUS AUREUS

AND STREPTOCOCCUS SPP. FROM CLINICALLY MASTITIC MILK SAMPLES

J. L. McCarron, G. P. Keefe, S. L. McKenna, I. R. Dohoo, and D. E. Poole

Published in J Dairy Sci 2009. 92 :5326-5333.

Revised Version June 2012

44 3.1. Abstract

The primary objective was to compare microbiological results of the University of

Minnesota Tri-plate and the 3M Petrifilm Staph Express (STX) Count Plate to standard culture techniques for identification of clinical mastitis causedStaphylococcus by aureus.

The secondary objective was to evaluate the Tri-plate’s ability to differentiateStreptococcus spp. from other gram positive organisms. The tests were evaluated using clinically positive mastitic milk samples (n = 282) to determine their ability to diagnose the pathogens of interest. A Tri-plate was classified positive forS. aureus when at least 1 colony exhibiting (5 hemolysis was present on the Factor medium portion of the plate. When used in this manner and read by a trained laboratory technician, the sensitivity of the Tri-plate was 97.9% (95%

Cl: 88.7 - 99.9) and the specificity was 81.8% (95% Cl: 75.9 - 86.7). ’When the Tri-plate was evaluated by the laboratory technician for its ability to diagnoseStreptococcus a spp., both the sensitivity and specificity of the test were very good (92.5% (95% Cl: 84.4 - 97.2) and 89.5% (95% Cl: 84.1 - 93.6), respectively). Using the Petrifilm, samples were classified as positive for S. aureus if any red-violet colonies were present on the Petrifilm after an initial 24 h incubation. When used in this manner the Petrifilm had a sensitivity of 97.4%

(95% Cl: 86.2 - 99.9) and a specificity of 76.1% (95% Cl: 68.7 — 82.5). Further evaluation of the Petrifilm was done using the STX Disk, which was used to confirm the presence S.of aureus. When using the presence of 1 pink colony on the disk, the sensitivity of the Petrifilm was 92.1% (95% Cl: 78.6 - 98.3) and the specificity was 93.1% (95% Cl: 88.0 - 96.5). Both the Tri-plate and the 3M STX Petrifilm successfully diagnosedS. aureus in clinical milk

45 samples when used in a laboratory setting and the Tri-plate successfully differentiated

Streptococcus spp. from other gram positive organisms.

3.2. Introduction

Mastitis is the most costly infectious disease in the dairy industry (1) and as a result much work has been done developing comprehensive control programs aimed at disease prevention. Even in situations where all preventive measures are carried out, there continues to be cases of clinical mastitis. The presence of these cases requires farms to incorporate diagnostic and treatment protocols into their overall mastitis control programs. These treatment protocols should be farm-specific and recognize the predominant pathogens (2).

Knowing the pathogens present on a farm, and in an individual cow, will allow producers to make rational therapy decisions as well as implement specific control measures that would be best suited for the pathogen present. Continued surveillance will allow producers detection of the presence of new or emerging pathogens on the farm (3).

Rational use of antibiotics and appropriately targeted treatment strategies have been the major foci of mastitis control programs (2). Yet, therapy of clinical mastitis remains a topic of debate as no consistent approach to the early identification and treatment of clinical mastitis cases has been developed. Ideally, aseptically acquired milk samples obtained from each clinical case of mastitis would be cultured in a laboratory prior to making an individual cow treatment decision. But, in many cases, and for various reasons, treatment decisions are made empirically. Leslie et al. identified the need for rapid, sensitive, specific and low cost aids to categorize cases of IMI (4). Since then, various studies evaluated cow-side tests able

46 to successfully categorize mastitis-causing pathogens into potential treatment categories (5,

6,7).

Two tests with the ability to rapidly categorize clinically mastitic milk samples into broad treatment categories are the University of Minnesota Tri-plate and the 3M Petrifilm.

Because of variation in growth requirements for some bacteria, on-farm culture systems could not effectively detect all pathogens associated with mastitis (8). The University of

Minnesota Tri-plate is divided into 3 types of media. Factor is specific for gram positive organisms; modified thallium sulfate-crystal violet - B toxin blood (MTKT), is specific for

Streptococcus spp.; and MacConkey, is specific for gram negative organisms. The Factor medium of the plate is primarily used for the identification ofaureus 5. andStaphylococcus spp. S. aureus appears as creamy, grayish or golden colonies with a zone of complete hemolysis around the colony (9). Testing for the presence and type of hemolysis on blood agar-based plates represented the first simple and rapid method for the detectionS. aureus of in milk samples (10). However, previous research showed that approximately 20 to 25% of the S. aureus isolates from bovine mastitis do not show detectable P hemolysis activity in primary cultures, making this a specific, but not very sensitive criterion for the identification of S. aureus (11).

The 3M Petrifilm Staph Express Plate (STX) is designed to be selective and differential forStaph, spp. The Petrifilm produces results within 24 h ± 2 h of incubation. A further 3 h of incubation is required, when the confirmatory STX Disk is applied (12). Rapid results may be preferred in situations where timely decisions are required, such as in on-farm culture programs used for treatment decisions (13). Many studies noted that sensitive and

47 rapid cow-side diagnostic kits for the identificationS. of aureus in milk could play an important role in the control of contagious mastitis (11, 14, 15).

The primary objective was to compare microbiological results of the University of

Minnesota Tri-plate and the 3M Petrifilm STX to standard culture techniques for identification ofS. aureus in clinically mastitic milk samples. The secondary objective was to evaluate the Tri-plate’s ability to differentiateStreptococcus spp. from other gram positive organisms.

33. Materials and Methods

Details of sample collection and processing, including laboratory procedures for gold standard testing were presented in Chapter 2. A brief summary of the procedures is presented below.

3.3.1. Samples

Samples of clinically mastitic milk (n = 282) were collected from 21 dairy farms in

Prince Edward Island. Samples with complete records (n = 271) were used for Tri-plate evaluation and a subset (n = 213) was used for evaluation of the Petrifilm STX plate.

Producers were trained to recognize abnormal (mastitic) milk and aseptically collect all milk samples. Samples were never frozen, kept refrigerated and transported to the Atlantic

Veterinary College for culture within 24 to 36 h of collection. As in Chapter 2, the 282 milk samples received were considered independent. The analysis presented does not take into consideration the hierarchical structure of the data, thus the effect of clustering was ignored.

The standard errors of Se, Sp, NPV and PPV should be interpreted with this in consideration.

48 3.3.2. Gold Standard

Gold standard bacteriological cultures were performed according to the Laboratory

Handbook on Bovine Mastitis (16). Samples were classified as having significant growth if the growth was considered of ‘probable significance’ or ‘highly significant’ based on

National Mastitis Council Guidelines for significance (16). Disposable plastic loops were used to streak 10 pL of each sample onto blood agar and MacConkey plates. Plates were incubated at 35 °C for 24 h. The standard laboratory plates were read by a milk laboratory technician and colonies were identified based on growth characteristics, morphology, pattern of hemolysis, catalase reaction and Gram-staining properties. The tube coagulase test was used to differentiateS. aureus from other coagulase negativeStaphylococcus species.

Samples that grew a yeast or mold were classified as no bacterial growth. Samples that had 2 colony types were considered mixed growth and samples with 3 or more were considered contaminated. If samples exhibited mixed growth or contamination on gold standard culture, they were not used in the evaluation of the Tri-plate and Petrifilm.

3.3.3. Minnesota Easy Culture System II Tri-plate

The Minnesota Easy Culture System II Tri-plate, developed by the University of

Minnesota’s Laboratory for Udder Health (St. Paul, Minnesota), is a culture plate that is divided in 3 sections. One section contains a proprietary Factor medium that is selective for gram positive bacteria, 1 section contains MacConkey medium for the identification of gram negative bacteria and 1 section contains MTKT medium that is selective Streptococcusfor species (9). The media were inoculated according to the manufacturer’s recommendations.

Sterile cotton tipped swabs were saturated in milk and used to inoculate each section of the

49 plate, re-dipping the swab in the sample before inoculating each section. Plates were incubated in a 35 °C incubator for 24 h before being read.

The criterion used to classify a sample as positive forS. aureus was growth on the

Factor medium that exhibited hemolysis around at least 1 colony. One colonyS. of aureus was considered significant by the National Mastitis Council, so plates were assigned toS. the aureus category if a single colony exhibiting hemolysis was present. If there was growth on the Factor medium that did not exhibit hemolysis, growth on either of the other media or no growth on any media, the sample was assigned to the aureus5. negative category. Each Tri­ plate was read by a trained milk microbiology technician as well as 4 masked readers with limited microbiology experience to determine their ability to detectS. aureus by identifying hemolysis on Factor media.

Tri-plates were also evaluated for their ability to diagnoseStrep, spp. Plates were considered positive forStreptococcus, spp. if there was colony growth (at least 1 colony) on the MTKT media. Again, each Tri-plate was read by a trained milk microbiology technician as well as 4 masked readers with limited microbiology experience.

3.3.4.3M Petrifilm Staph Express

The second media system used was the 3M Petrifilm STX. The Petrifilm STX is a sample-ready culture medium system, which contains a coldwater-soluble gelling agent. The chromogenic, modified Baird-Parker medium in the plate is selective and differentialS. for aureus. Red-violet colonies on the plate areS. aureus (12). In cases where the color of the colonies was not easily identified or when colonies other than red-violet were present on the plate, the 3M Petrifilm STX disk may be used to identifyS. aureus. The STX disk contained

50 a dye and DNA.S. aureus produces deoxyribonuclease that reacts with the dye to form pink zones. Milk samples were diluted 1:10 with sterile water, a lmL aliquot was plated on each

STX Petrifilm and plates were incubated at 35 °C for 24 h. Colony growth, number of colonies present and color of colonies were recorded by the laboratory technician. Initial test characteristics (sensitivity, specificity and predictive values) were calculated based solely on the presence of red-violet colonies as per the manufacturer’s interpretation criteria.

Subsequently, all Petrifilms that were positive for growth (at least 1 colony) had the STX disk inserted into the plate and were re-incubated at 37° C for 3 h (n = 123). Plates that exhibited at least 1 colony with a pink zone were classified positive forS. aureus. Pink colonies were then picked and re-grown using standard methods to confirm their identity.

All Petrifilms were read by the 4 masked readers and the laboratory technician; however, after the STX disks were applied, plates were read only by the laboratory technician; therefore, no inter-reader comparisons were made.

3.3.5. Statistical Analysis

All results were analyzed using Intercooled Stata 9(17). For assessment of the Tri­ plate systems’ ability to identifyS. aureus andStrep, spp., sensitivity (Se), specificity (Sp), positive predictive value (PPV) and negative predictive value (NPV) were calculated by comparing the Tri-plate classifications to the gold standard results. The Petrifilm STX was evaluated in the same manner to determine its ability to diagnoseS. aureus. Agreement between the 4 masked readers was determined by calculating kappa statistics.

3.4. Results

3.4.1 Samples

51 A total number of 282 fresh milk samples were received from producers. Three samples had incomplete records and 18 samples were removed from the analysis because they exhibited mixed or contaminated growth on the gold standard, leaving 261 samples for

Tri-plate analysis. Gold standard culture results for each sample used in the analysis are in

Table 3.1. The prevalence ofS. aureus used to evaluate the Tri-plates was 18.0% and the prevalence of all Strep, spp. combined was 30.7%.

There were 213 samples that had complete records on the Petrifilm STX. After 6 wk of sample collection, the laboratory technician noted that lmL of undiluted milk plated on the Petrifilms resulted in some plates that were difficult to interpret (milk clots and cases of heavy bacterial growth resulted in difficulty identifying individual colonies). At that point, a second series of Petrifilms with a 1 in 10 dilution (equivalent to 100 pL of milk and 900 pL of sterile diluent on the film) was included. There were 64 samples received at the beginning of the collection period that were not diluted, leaving 213 samples with complete records.

Sixteen were not included in the final analysis, because they were contaminated on gold standard culture. Previous research evaluating the Aerobic Count and Coliform Count

Petrifilms’ performance using diluted vs. undiluted milk showed no difference in Se and

NPV, but slightly higher Sp and PPV when diluted milk was used (See Chapter 2). Wallace et al. found test characteristics of the STX were the highest for diluted (1 in 10) fresh samples from clinically mastitic cows (18). For these reasons, only diluted milk samples were included in the analysis. Table 3.1 illustrates the gold standard culture results for each of the samples used to determine test characteristics of the STX Petrifilm. The prevalence of

S. aureus in the samples used to evaluate the Petrifilm was 19.3%.

52 3.4.2 Minnesota Easy Culture System II Tri-plate

Table 3.2 illustrates the test characteristics of the Tri-plate when read by the laboratory technician. The Tri-plate was highly sensitive because only 1 sample that was positive for S. aureus on gold standard was not diagnosed using the Tri-plate. The reduced

Sp was the result of 39 samples being classified asS. aureus by the Tri-plate that were not diagnosed asS. aureus by the gold standard methods. Most (12) of these samples had no bacterial growth on the gold standard, the remaining samples includedS. uberis, 9 7 Strep, spp, 5 S. dysgalactiae, 5 E. coli and 1Staph, spp.

Table 3.3 shows the test characteristics of the Tri-plates when read by the 4 masked readers who used a threshold of 1 colony that exhibited p hemolysis on Factor media to consider a plate positive forS. aureus. The Se of the Tri-plate was lower when read by the readers (43.2 to 59.1%) with limited microbiology training than when read by the technician.

The level of agreement among the readers was moderate, with the calculated kappa value

(actual agreement beyond chance) being 0.51. The Sp of the test was higher with the inexperienced readers, ranging from 93.8% to 95.9%. All test characteristics calculated were similar between the readers.

Table 3.4 shows the test characteristics of the Tri-plate when evaluated by the laboratory technician for its ability to diagnose Strep,a spp. Sensitivity and Sp of the test were 92.6% and 89.5% respectively. There were 6 samples that were diagnosed asStrep. spp. using the gold standard that did not grow on the MTKT media. There were 19 samples that exhibited growth on the MTKT media that were not diagnosedStrep, as spp. using gold standard. On gold standard, 9 of the 19 were diagnosed no growth,S. aureus, 5 3 E. coli, 1

53 Corynebacterium spp. and 1Staph, spp. The same masked readers evaluated the Tri-plate for the presence ofStrep spp. The sensitivities of the Tri-plate, when read by these readers, ranged from 77.8% to 88.9%, and the specificities ranged from 83.7% to 92.7%, with a combined kappa value of 0.81.

3.4.3 3M Petrifilm Staph Express

Samples were classified as positive forS. aureus if any red-violet colonies were present on the Petrifilm after an initial 24 h incubation. Test characteristics and predictive values are presented in Table 3.5. When used in this manner the Petrifilm was highly sensitive (97.4%) but not very specific (76.1%). Only 1 sample that was diagnosedS. aureus using the gold standard did not show growth of red-violet colonies. There were 38 samples that grew red-violet colonies, but were not classifiedS. aureus using the gold standard culture. Eight of these samples showed no growth on gold standard culture, 16 wereStrep. spp., 7 were Staph, spp., 6 were gram negative coliforms and 1 wasCorynebacterium bovis.

Table 3.5 illustrates the test characteristics of the Petrifilm STX after the application of the STX disk and re-incubation for 3 h. All plates were read only by the laboratory technician and the presence of 1 pink colony was used to classify a sample as positive S.for aureus. The Petrifilm STX with disk applied was highly sensitive (92.1%) because only 3 of the 38 samples that were positive for S. aureus on gold standard were not detected using the

STX Petrifilm. Two of the 3 samples did not show any visible growth on the Petrifilm and the remaining sample did show colony growth, but was not confirmed pink using the disk.

The Sp of the test was increased when the STX disk was used (from 76.1 % to 93.1 %). This was the result of only 11 samples being classified asS. aureus by the Petrifilm STX with disk

54 applied that were not diagnosed S.as aureus using the gold standard. Each of the 11 samples had the pink colonies picked from the plate and re-cultured using standard methods. Five were confirmedS. aureus, 4 were Staph, spp., 1 was S. hyicus, and 1 yielded no growth. Had the 5 samples that were diagnosed asS. aureus on subsequent standard culture methods been diagnosed by the original gold standard culture, the Se of the Petrifilm would be increased to

93.0% and the PPV would increase to 87.0%.

When based on results of the STX Petrifilm using the presence of red-violet colonies to diagnoseS. aureus, the apparent prevalence was 38.1% and when the diagnosis was based on confirmation by the STX disk, the apparent prevalence ofS. aureus was 23.4%. The application of the disk did not result in any new samples being correctly identifiedS. as aureus positive.

35. Discussion

A complete mastitis control program should include the use of microbiologic analysis of individual milk samples to determine which mastitis pathogens are present on the farm (5).

Having on-farm access to accurate, rapid microbiological tests that are capable of diagnosing some of the most prevalent mastitis pathogens may facilitate the incorporation of this valuable data into mastitis control programs.

When implementing preventative strategies for mastitis, information about the prevalence of specific pathogens is useful (19). The National Mastitis Council

Recommended Mastitis Control Program suggests milking cows with confirmed contagious

IMI last, and marketing or permanently segregating cows that are persistently infected with

S. aureus or other non-responsive microbial agents. Confirming the presenceS. of aureus

55 quickly and accurately may allow better execution of segregation strategies. For producers making decisions on treatment protocols (routine, extended therapy, withholding treatment and culling) identificationS. of aureus is of value (20).

The 2 mastitis-causing pathogens that were the focus here areS. aureus andStrep. spp. Both of these pathogens are considered significant mastitis causing organisms by the

National Mastitis Council (16) and are 2 of the most prevalent pathogens encountered. For epidemiologic monitoring and investigation of risk factors for prevention of cases, identification of the type of gram positive organism(Strep, spp. or S. aureus) is important.

Strep, spp. were isolated from 12.2% of milk samples submitted to the Wisconsin Veterinary

Diagnostic Laboratory, andS. aureus was isolated from 9.7% of samples (19). More recently, a Canadian prevalence study of clinical mastitis found 12.5% of samples submitted were positive for Strep, spp and 10.3% were positive forS. aureus (21). Strep, spp. were the most prevalent species cultured. Of the 261 samples used to evaluate the Tri-plate, 30.7% grew Strep, spp. on gold standard culture.S. aureus was the second most prevalent organism cultured on gold standard, with a prevalence of 18.0%. In the 197 samples used to evaluate the Petrifilm, 30.5% were diagnosedStrep, spp. on gold standard and 19.3% were diagnosed as S. aureus. '

The Tri-plate culture system was designed to identify some of the most common pathogens infecting the bovine udder (22). Goddenet al. evaluated the system in laboratory and field studies for its ability to differentiate growth from no growth, gram positive from gram negative and growth ofStaph vs. Strep spp. They found the system accurate, and it attained a high level of agreement (kappa values between 0.80 and 0.93) when used for those

56 3 purposes. But, when users tried to differentiate pathogen groups further or to identify specific pathogens (S. aureus vs. Coagulase Negative Staph spp.) much lower agreement was observed (8). When the University of Minnesota Bi-plate (which uses Factor and

MacConkey media) was evaluated for its ability to differentiate growth from no growth and gram positive from gram negative organisms, the test was highly sensitive (97.9%) (Chapter

2). Similarly, agreement among readers was high (kappa = 0.76) when asked to differentiate the above categories. In the current research, when readers were asked to identifyS. aureus on the Tri-plate, in concurrence with Goddenet al., the agreement among readers was low

(kappa = 0.51) (8).

Beta hemolysis represents an important criterion for rapid presumptive identification of S. aureus in primary cultures (11). Of the 3 coagulase positiveStaph, spp. regularly encountered in clinical milk samples,S. aureus was the only one with double zone hemolysis

(10). Yet, 20 to 25% of the S. aureus isolates from bovine mastitis did not present detectable hemolysis in primary cultures. The proportion of hemolyticS. aureus found in bovine mastitis varies from region to region (23). Therefore, the performance of a test that uses hemolysis as part of the diagnostic criteria may vary with regards to the population under investigation (11). In this study, the Tri-plate was highly sensitive (97.9%) with moderate Sp

(81.8%) when used by a laboratory technician to diagnose the presenceS. ofaureus. But, when read by masked readers, the sensitivities were much lower (ranging from 43.2% to

59.1%). In contrast, the Sp of the test was higher for the inexperienced readers, ranging from

93.8% to 95.9%, possibly because these readers were more reluctant to call a colony positive resulting in fewer false positives. Similar work by Lam et al. used the presence or absence of

57 P hemolysin production as a method to diagnoseS. aureus (10). Plates were read by an experienced observer and 45 out of 54 samples were correctly identified, resulting in a Se of

83% and a Sp of 98%. Predictive values of the test were calculated at various levels of prevalence. At a prevalence of 21% (the median prevalence among the herds in their study) the PPV was 91% and the NPV was 96%.

When used by the inexperienced readers, observation of hemolysis resulted in a highly specific test with very high negative predictive values. Farm-based users of the Tri­ plate in this manner could be confident in detecting animals that are negativeS. aureus for and depending on the prevalence in the herd, could be very confident in a negative test result.

Unfortunately, if this test were used to detect positive cows for segregation from the herd, a user without extensive microbiological experience would miss a number of positive animals.

The very high Se obtained by the laboratory technician indicates that with thorough training, the Tri-plate can be used to identify animals truly infected withS. aureus in a farm situation.

Tri-plates were evaluated for their ability to diagnoseStreptococcus, spp. based on the presence of colony growth on the MTKT media. When read by the technician both the

Se and Sp of the test were high (92.6% and 89.5%, respectively). The lower Sp was the result of 19 samples showing growth on the MTKT media that were not diagnosed as

Streptococcus, spp. on gold standard. Almost half (9) of the samples showed no growth on gold standard. This may be due to a larger innoculum of milk used on the Tri-plate (one saturated cotton swab for one-third of the plate vs. the 10 pL loop used for gold standard).

There were 5 samples that were diagnosed asS. aureus on gold standard that exhibited

58 colony growth on MTKT. This can be explained by the presence of 1 colony S.of aureus in the gold standard being diagnosedS. aureus.as

The Petrifilm STX plate was designed to provide rapid results for the diagnosisS. of aureus after 24 ± 2 h of incubation when red-violet colonies are present. The Petrifilm STX

Disk should be applied to the plate and re-incubated for 3 h whenever colonies other than red-violet are present on the plate; for example, black or blue-green colonies (12). The manufacturer’s interpretive criteria suggest that the appearance of red-violet colonies on the initial incubation is presumptive evidence for the diagnosisS. ofaureus. Results from Silva et al. did not support this recommendation and found it necessary to use the Staph Express

Disk for confirmation even when red-violet colonies were the only ones present (13). They found that the apparent prevalence ofS. aureus in milk samples processed using Petrifilm was significantly greater than the prevalence in milk samples processed using standard microbiological techniques. The results of the current study show that the STX Petrifilm had a higher Se when using only the presence of red-violet colonies than when the presence of pink colonies on the Staph Express Disk was used, 97.4 vs. 92.1%. In the population of samples that were used in this study, there were only 38 that were diagnosedS. aureus on gold standard. The difference in sensitivities was due to the difference in classification of only 2 samples; therefore, this difference may be of little biological significance. The high sensitivities obtained in both methods of evaluation indicate that using the STX Petrifilm resulted in very few false negative classifications. This is an important attribute for a test to have if it were being used as part of a mastitis control program as the occurrence of false negative results could result in the maintenance of infected animals within a herd (13). The

59 Sp of the STX Petrifilm was greatly improved by the use of the STX disk (from 76.1% to

93.1%). Results from Silva et al. showed that the STX Petrifilm was highly specific (98.5%)

when the disk was used (13). In the current study, the Sp of using the presence of red-violet

colonies was decreased because there were 38 samples that were diagnosedS. aureus using

the Petrifilm that were notS. aureus on gold standard. These results show that a variety of

organisms other thanS. aureus will grow red-violet on the STX Petrifilm; therefore, caution

should be used when using this test without the confirmatory disk, especially if segregation

and or culling decisions are made based on the results. The application of the STX disk did

not improve the Se of the test as the manufacturer suggests, rather the Sp of the test was

greatly improved.

In the case of hemolysis on the Tri-plate, where previous experience of the reader

would potentially affect the outcome of the test, readers with extensive experience and

inexperienced readers were used. Our findings, along with results of other studies that have

evaluated potential on-farm tests for mastitis, support the need for thorough training of

individuals reading the tests. Previously published work also recommends that periodic

assessment of accuracy of on-farm methods by submission of duplicate samples to a

microbiology laboratory be carried out (13).

When the 2 tests were read by the laboratory technician, there was no difference in

the detection of 5. aureus (Se). The specificities of the 2 tests were similar when the

presence of red-violet colonies only was used to assess the Petrifilm. But, with the

. application of the STX disk, the Petrifilm was significantly more specific test than the Tri­

plate.

60 3.6. Conclusion

Both the University of Minnesota Tri-plate and the 3M STX Petrifilm were able to successfully detectS. aureus in clinically mastitic milk samples when used in a laboratory setting. The presence of P hemolysis on the Tri-plate was a highly sensitive method to diagnoseS. aureus when read by a trained laboratory technician. Yet, when read by individuals with limited microbiology experience, the Se of the test was much lower.

Specificity of the Tri-plate system was higher when read by the inexperienced readers than by the laboratory technician. The Tri-plate was able to successfully differentiateStrep, spp. from other gram positive organisms.

The 3M Petrifilm was very sensitive when used to diagnose the presenceS. of aureus.

When using the presence of red-violet colonies only to diagnose a sample positive withS. aureus, the Se and Sp of the Petrifilm were similar to those of the Tri-plate. The Sp of the

Petrifilm was increased greatly by using the STX confirmatory disk.

To determine their suitability as on-farm tests for specific mastitis causing pathogens, each of these tests should be evaluated in field situations. Training of individuals performing and reading the tests will be key to their success on the farm. For quality control purposes, samples taken on-farm should be saved and submitted to a diagnostic laboratory from time- to-time to assess the on-going accuracy of any on-farm culture system.

61 3.7. Acknowledgements

The authors thank Theresa Andrews, Lloyd Dalziel, Katie Macintosh, Michael

Trenholm, Cynthia Mitchell, and Shana Richard of the Department of Health Management for their technical help with this project

3.8. References

1. Erskine, R. J., S. Wagner and F. J. DeGraves. 2003. Mastitis therapy and pharmacology. Vet. Clin. North Am. Food Anim. Pract. 19:109-138.

2. LeBlanc, S. J., K. D. Lissemore, D. F. Kelton, T. F. Duffield and K. E. Leslie. 2006. Major advances in disease prevention in dairy cattle. J. Dairy Sci. 89:1267-1279.

3. Ruegg, P. L. 2003. Investigation of mastitis problems on-farms. Vet. Clin. North Am. Food Anim. Pract. 19:47-73.

4. Leslie, K. E., J. T. Jansen and G. H. Lim. 2002. Opportunities and implications for improved on-farm cowside diagnostics. Page 147-160 in Proc. DeLaval Hygiene Symposium, Guelph, ON.

5. Ruegg, P. 2005. On-farm diagnosis of mastitis infections and organisms. Page 24-30 in Natl. Mast. Counc. Reg. Mtg. Proc, Burlington, VT.

6. Lago, A., S. Godden, R. Bey, K. Leslie, R. Dingwell and P. Ruegg. 2006. Validation of the Minnesota Easy Culture System II: Results from on-farm bi-plate culture versus standard laboratory culture. Page 250-251 in Am. Assoc.Bov. Prac. Ann. Mtg. Proc., St.Paul, MN.

7. McCarron, J. L., G. P. Keefe, S. L. McKenna, I. R. Dohoo and D. E. Poole. 2009. Laboratory evaluation of 3M Petrifilms and University of Minnesota Bi-plates as potential on-farm tests for clinical mastitis. J. Dairy Sci. 92:2297-2305.

8. Godden, S., A. Lago, R. Bey, K. Leslie, P. Ruegg and R. Dingwell. 2007. Use of on- farm culture systems in mastitis control programs. Page 1-9 in Natl. Mastitis Counc. Reg. Mtg. Proc., Visalia, CA.

9. Laboratory for Udder Health, Minnesota Veterinary Diagnostic Laboratory. 2000. Minnesota Easy Culture System II Handbook. University of Minnesota, St. Paul, MN.

62 10. Lam, T. J. G. M., A. Pengov, Y. H. Schukken, J. A. H. Smit and A. Brand. 1995. The differentiation of Staphylococcus aureus from other micrococcaceae isolated from bovine mammary glands. J. App. Bacteriol. 79:69-72. Paul, MN.

11. Boerlin, P., P. Kuhnert, D. Hussy and M. Schaellibaum. 2003. Methods for identification of Staphylococcus aureus isolates in cases of bovine mastitis. J. Clin. Microbiol. 41:767-771.

12. 3M Microbiology. 2005. 3M Petrifilm Interpretation Guide. 3M Microbiology, St. Paul, MN.

13. Silva, B. O., D. Z. Caraviello, A. C. Rodrigues and P. L. Ruegg. 2005. Evaluation of Petrifilm for the isolation of Staphylococcus aureus from milk samples. J. Dairy Sci. 88:3000-3008.

14. Hogan, J., A. Cometta and J. Pankey. 1986. Comparison of four test procedures to identify Staphylococcus aureus isolated from bovine intramammary infections. Am. J. Vet. Res. 47:2017-2019.

15. Watts, J. and W. Owens. 1988. Evaluation of the rapid mastitis test for identification of Staphylococcus aureus and Streptococcus agalactiae isolated from bovine mammary glands. J. of Clin. Microbiol. 26:672-674.

16. National Mastitis Council. 1987. Laboratory and Field Handbook on Bovine Mastitis. 1st ed. W. D. Hoard and Sons Co., Fort Atkinson, WI.

17. StataCorp. 2005. Intercooled Stata 9.0. College Station, Texas, USA.

18. Wallace, J., J. Roy, E. Bouchard, L. DesCoteaux, S. Messier and D. DuTremblay. 2008. Comparison of 3M Petrifilm Staph Express Count plates, 3M Petrifilm Rapid Coliform Count plates and 3M Aerobic Count plates with standard bacteriology of bovine milk. Page 162-163 in Natl. Mast. Counc. Ann. Mtg., New Orleans, LA.

19. Makovec, J. A. and P. L. Ruegg. 2003. Results of milk samples submitted for microbiological examination in Wisconsin from 1994 to 2001. J. Dairy Sci. 86:3466-3472.

20. Sol, J., O. C. Sampimon, H. W. Barkema and Y. H. Schukken. 2000. Factors associated with cure after therapy of clinical mastitis caused by Staphylococcus aureus. J.Dairy Sci. 83:278-284.

21. Olde Riekerink, R., H. Barkema, D. Kelton and D. Scholl. 2008. Incidence rate of clinical mastitis on Canadian dairy farms. J. Dairy Sci. 91:1366-1377.

63 22. Bey, R. and R. Farnsworth. 2000. Cow side microbiology. Page 97 in Proceedings of the Annual Meeting of the Minnesota Dairy Health Management Conference, St.

23. Larsen, H., F. Aarestrup and N. Jensen. 2002. Geographical variation in the prevalence of genes encoding superantigenic exotoxins and P-hemolysin among Staphylococcus aureus isolated from bovine mastitis in Europe and USA. Yet. Microbiol. 85:61-67.

64 Table 3.1. Gold standard microbiology results of samples used to evaluate the University of Minnesota Tri-plate and the 3M Petrifilm Staph Express______Gold Standard Result Samples used for % Samples used for % Tri-Plate Evaluation Petrifilm Evaluation No Bacterial Growth 85 32.6 61 30.1 Staphylococcus aureus 47 18.0 38 19.3 Streptococcus uberis 43 16.5 36 18.3 Escherichia coli 27 10.3 18 9.1 Streptococcus spp. 18 6.9 11 5.6 Streptococcus dysgalactiae 18 6.9 13 6.6 Coagulase Negative Staph 6 2.3 5 2.5 spp. Staphylococcus spp. 4 1.5 4 2.0 Pasteurella multocida 3 1.2 3 1.5 Staphylococcus hyicus 3 1.2 3 1.5 Klebsiella spp. 2 0.8 2 1.0 Streptococcus bovis 1 0.4 0 0 Bacillus spp. 1 0.4 0 0 Corynebacterium spp. 1 0.4 1 0.5 Serratia spp. 1 0.4 1 0.5 Enterobacter spp. 1 0.4 1 0.5 Total 261 100 197 100

65 Table 3.2. Test characteristics of the University of Minnesota Tri-plate when read by an experienced technician based on its ability to correctly classify samplesS. aureus as (n=271)

Test Characteristic (95% Confidence Interval) Sensitivity 97.9% (88.7-99.9) Specificity 81.8% (75.9 - 86.7) Positive Predictive Value 54.1% (43.0-65.0) Negative Predictive Value 99.4% (96.9 - 100)

6 6 Table 3.3. Test characteristics of the University of Minnesota Tri-plate based on its ability to correctly classify samples as S. aureus when read by masked readers with limited microbiology training

Test Characteristic Reader A Reader B Reader C Reader D Sensitivity 51.1% 43.2% 54.6% 59.1% Specificity 93.8% 95.2% 94.8% 95.9% Positive Predictive Value 63.9% 65.5% 68.6% 76.5% Negative Predictive Value 90.0% 88.8% 90.9% 91.3%

67 Table 3.4. Test characteristics of the University of Minnesota Tri-plate when read by an experienced technician based on its ability to correctly classify samplesStrep as spp. (n=271)

Test Characteristic (95% Confidence Interval) Sensitivity 92.5% (84.4-97.2) Specificity 89.5% (84.1-93.6) Positive Predictive Value 79.8% (69.9-87.2) Negative Predictive Value 96.4% (92.4-98.7)

6 8 Table 3.5. Test characteristics of the 3M Staph Express Petrifilm when read by an experienced technician using the presence of red-violet colonies on initial culture to classify samples asS. aureus vs. using the confirmatory disk

Test Characteristic (95% Red Violet Colonies Staph Express Disk Confidence Interval) Sensitivity 97.4% (86.2-99.9) 92.1% (78.6- 98.3) Specificity 76.1% (68.7-82.5) 93.1% (88.0- 96.5) Positive Predictive Value 49.3% (37.6-61.1) 76.1% (61.2- • 87.4) Negative Predictive Value 99.2% (95.5 - 100) 98.0% (94.3- 99.6)

69 CHAPTER 4. EVALUATION OF 3M PETRBFILM, UNIVERSITY OF MINNESOTA

TRI-PLATE AND SOMATIC CELL COUNT AS DIAGNOSTIC TESTS USED TO

DETERMINE INFECTION STATUS AT TIME OF DRY-OFF

70 4.1. Abstract

The objective of this study was to evaluate the cow level test characteristics of two on-farm tests with the potential to be used as part of a selective dry cow therapy program.

The 3M Petrifilm (1) and the University of Minnesota Tri-plate (2) were evaluated for their ability to classify a cow as either infected or uninfected at the time of dry-off. Each test was evaluated individually, as well as in combination with somatic cell count (SCC) data obtained from monthly Dairy Herd Improvement (DHI) records. The study used 820 quarter samples collected from Holstein cattle on 18 dairy herds in Atlantic Canada. Individual quarter samples were collected by producers at the time of dry-off and submitted for bacterial culture. Composite SCC data collected for DHI from three months prior to dry-off were also obtained from each cow in the study. A range of definitions for infection was created using the results from the above tests. When the National Mastitis Council criteria of significance were applied to the gold standard culture results, 11.9% of samples received were classified as having microorganisms of probable or high significance. Using the 3M Petrifilm alone to classify a sample as infected resulted in a sensitivity of 100% and a specificity of 61.2%.

Using the Tri-plate alone resulted in a sensitivity of 96.4% and a specificity of 44.6%. Using only the last SCC prior to dry off resulted in sensitivities of 50.2% and 36.9% (cut-offs of

150 000 and 200 000 respectively). When Tri-plate results were combined with SCC data, the sensitivity of the test was improved to 98.5%, however, the specificity of the test was decreased to 31.7% and 34.2%, using 150 000 and 200 000 as SCC cut offs, respectively.

These results suggest that both the Petrifilm and the Tri-plate are sensitive tests that could be used as part of a selective dry cow treatment program.

71 4.2. Introduction

Management of mastitis on dairy farms continues to emphasize prevention rather than treatment. One key area of mastitis prevention is the dry period, where the prevention of new infections as well as the treatment of existing infections have been traditionally accomplished by blanket antimicrobial dry cow therapy. Gram positive infections, most commonly,

Staphylococcus aureus andStreptococcus uberis, are significant causes of existing intramammary infections at the time of dry-off in many dairy herds (3). In situations where cows are infected at the time of dry-off, the objective of the dry cow management program should be to cure the existing infection. When cows are uninfected at the time of dry-off, the goal of dry cow management is prevention of new infections from occurring (3). By using information about the infection status of a cow at the time of dry-off, a rational approach to dry cow therapy can be employed.

Public concern over the use of antimicrobials in animal agriculture requires the prudent use of antimicrobial dry cow therapy on dairy farms. Concerns that their use may promote bacterial antibiotic resistance and leave residues in the food chain have been raised

(4). In approximately 80% of cases, antibiotic residues in milk can be traced back to mastitis treatments given during lactation or the dry period (5). These concerns have prompted researchers to look at ways to selectively target dry cow therapy to animals that will benefit most from treatment. The prudent use of long acting dry cow therapy would involve treating only infected cows that are expected to respond to antibiotics and leaving uninfected cows without treatment (6). As uninfected cows are still at risk of developing a new infection during the dry period, producers may choose to use non-antibiotic methods of prevention,

72 namely internal teat sealants (ITS). In 2006, Sanfordet al. did not identify any significant treatment effect when use of an ITS alone was compared with use of cloxacillin alone in cows without evidence of infection late in the lactation period (7).

A selective dry cow therapy program requires the ability to identify cows that are most likely to benefit from treatment with antibiotics at the time of dry-off. Practically, a diagnostic test used for this purpose should be readily available on the farm, easy to use and interpret, and inexpensive. In 2006, Sanfordet al. reported that the California Mastitis Test had a sensitivity of 70% when used to detect mastitis causing pathogens at the time of dry-off

(8). In 2008, a study done by Torreset al. evaluated the use of clinical mastitis history and somatic cell counts (SCCs) from monthly Dairy Herd Improvement (DHI) records for the identification of infected and uninfected cows at dry-off. Various thresholds of cell counts were used to classify cows as infected, with resulting sensitivities calculated between 62.5% and 85.1% (6). Somatic cells are elevated when an inflammatory response occurs in the udder, therefore, they are commonly used to distinguish between infected and uninfected quarters (5). Somatic cell count data are readily available to dairy producers that participate in DHI programs, making it a feasible tool that could be used in selecting cows that would benefit most from dry cow antimicrobial products. Torreset al. concluded that cows with intramammary infections at the time of dry-off could be adequately identified by combining information from somatic cell count and clinical mastitis history but cautioned that decisions regarding selection criteria and adaption of selective dry cow therapy depend on the prevalence of intramammary infections in a herd and the type of microorganisms involved

(6).

73 Microbiology is considered the gold standard test for determining the infection status of a cow (8). Traditionally, microbiology-based tests have not been available on the farm, and cultures have not been routinely performed at dry-off due to logistic and financial constraints (9). Recent studies have identified various on-farm microbiologic tests that are able to successfully differentiate infected from uninfected cows during lactation (10, 11,12).

The objective of this study was to evaluate the ability of two on-farm tests, the 3M Petrifilm

(1) and the University of Minnesota Tri-plate (2), to classify a cow as either infected or uninfected at the time of dry-off. Each test was evaluated individually, as well as in combination with somatic cell count data obtained from monthly DHI records.

4 3. Materials and Methods

4.3.1. Samples

Two sets of milk samples were used for this study. Pre-dry-off quarter samples were obtained by dairy producers in Atlantic Canada (n=18) who were involved in the Canadian

Bovine Mastitis Research Network. A total of 820 quarter samples (from 206 cows) with complete records were received. Producers were instructed to aseptically collect quarter milk samples from 15 cows within one week of dry-off. In 2007, the first 15 cows to be dried off that were expected to remain in the herd until at least 2 wk after calving were sampled. Each sample was frozen on-farm and then shipped to the Atlantic Veterinary College for bacterial culture.

The second set of samples consisted of composite milk samples. All cows in the study were required to be enrolled in a monthly DHI program which included monthly collection of composite milk samples throughout the lactation. Sampling took place between

74 January 2008 and July 2008. These samples were used to determine the somatic cell count of the cows during the study period.

4.3.2. Gold Standard Culture

Gold standard bacteriological cultures were performed on the quarter samples according to the Laboratory Handbook on Bovine Mastitis (13). These samples were classified as having significant growth if the growth was considered to be of ‘probable significance’ or ‘highly significant’ based on National Mastitis Council Guidelines for significance (14). A cow was considered infected if any one of the quarter samples exhibited growth that was of probable or high significance according to the NMC guidelines. Frozen samples were thawed overnight at room temperature prior to culture. Samples were then mixed using a vortex mixer and disposable plastic loops were used to streak 10 pL of each sample onto blood agar and MacConkey plates. Plates were incubated at 35 °C for 24 h. The standard laboratory plates were read by a milk laboratory technologist and colonies were identified based on growth characteristics, morphology, pattern of hemolysis, catalase reaction, and Gram-staining properties. The tube coagulase test was used to differentiateS. aureus from other CNS species. Both primary and secondary species (if present) were reported. As per NMC protocol (4), samples with greater than 2 isolates were considered contaminated.

4.3.3.3M Petrifilm

Each milk sample was plated on a series of three Petrifilms, Aerobic Count (AC),

Coliform Count (CC) and the Staph Express (STX). The AC Petrifilm plate is a ready made culture medium that contains Standard Methods nutrients, a cold-water gelling agent and an

75 indicator dye that facilitates colony counting (1). It is used for counting aerobic bacteria.

The CC Petrifilm contains Violet Red Bile nutrients, a gelling agent and an indicator dye that facilitates colony counting. The Petrifilm STX also contains a coldwater-soluble gelling agent and a chromogenic, modified Baird-Parker medium that is selective and differential for

S. aureus (1). In cases where the colour of the colonies is not easily identified or when colonies other than red-violet are present on the STX plate, the instructions recommend that

3M Petrifilm Staph Express Disk may be used to identify S.aureus. The Staph Express disk contains a dye and deoxyribonucleic acid.S. aureus produces deoxyribonuclease (DNase) and the DNase reacts with the dye to form pink zones (1). Use of the Staph Express disk has been shown to improve the specificity of the test when used with clinically mastitic milk samples (15) (Chapter 3).

Milk samples were diluted 1:10 with sterile water (100 pi of milk in 900 pi of water) and plated on each of the three Petrifilms. All plates were incubated at 35 °C for 24 hours.

Petrifilm STX plates that were positive for growth (at least one colony) had the STX disk inserted into the plate and were re-incubated at 35 °C for l-3h. Plates that exhibited at least one colony with a clear pink zone were classified positive forS. aureus.

All AC and CC Petrifilms were read using the automated 3M Petrifilm Plate Reader.

STX Petrifilms were read manually by a trained laboratory technologist. Very few samples grew colonies on the CC or STX Petrifilm therefore only descriptive analysis of the STX and

CC Petrifilm results are presented. Results from the AC Petrifilm were used to create a definition of whether or not a quarter sample was positive for bacterial growth. As with the

76 Gold Standard, if any of the quarter samples from a cow were positive, the infection status of the cow was considered positive.

In order to consider a quarter sample positive, a colony count cut point of 5 on the AC plate was used. Previous research by McCarronet al. 2009 (12) (Chapter 2) showed that this cut point optimized test characteristics. Using the AC plate alone to define a sample as positive was a modification from previous work presented in Chapter 2. Coliform infections are rarely present at dry-off (16). The economic impact of treating these infections would be very little, therefore they were grouped with the AC positive samples. The presence of a single pink colony on the STX Petrifilm (confirmed with STX disk) was used to consider a sample positive for S. aureus.

4.3.4. Minnesota Easy Culture System II Tri-plate

The Minnesota Easy Culture System II Tri-plate, developed by the University of

Minnesota’s Laboratory for Udder Health, is a culture plate that is divided in three sections.

One section contains a proprietary Factor medium that is selective for gram positive bacteria, one section contains MacConkey medium for the identification of gram negative bacteria and one section contains MTKT medium that is selective forStreptococcus species (2). The media were inoculated according to the manufacturer’s recommendations. Sterile cotton tipped swabs were saturated in milk and used to swab each section of the plate, re-dipping the swab in the sample before swabbing each section. Plates were incubated in a 35 °C incubator for 24 h before being read.S. aureus was identified when there was growth on the

Factor media that exhibited p hemolysis around at least one colony. Each Tri-plate was read by a trained milk microbiology technologist and the criterion used to classify a sample as

77 positive for bacterial growth using the Tri-plate was growth of a single colony on any of the three sections of the plate. As with the Petrifilms, quarter samples were used for the analysis of the Tri-plate and if any of the quarter samples from one cow was determined to be positive, the infection status of the cow was deemed to be positive.

4.3.5. Somatic Cell Count

The DHI database was used to collect somatic cell count data for each cow in the study. Composite samples were collected on-farms by trained DHI technicians. A bronopol tablet (2-bromo-2-nitro-propane-l, 3 diol: 6 mg/tablet) (D & F Control Systems, Dublin, CA) was added to each sample vial to preserve the milk prior to somatic cell counting. The SCC of each quarter sample was obtained using the Fossomatic 4000 Milk Analyzer (Foss

Electric, Brampton, ON).

4.3.6. Statistical Analysis

All results were analyzed with Stata 9 (17). Using results from the two culture methods and SCC data, a variety of definitions were created to define a cow as infected (see

Table 4.1). Definitions were based on SCC (using cut-offs of either 150 or 200), Petrifilm or

Tri-plate culture results, or a combination of both. For example, for definition D, a cow was considered infected if her SCC was >150 (200) on her last test prior to dry-off or was positive on the Petrifilm, whereas definition C was only based on Petrifilm. Cow level definitions involving SCCs were created by using composite samples collected by DHI. Data obtained from the quarter level samples used on the Petrifilm and Tri-plates were aggregated to the cow level so that each cow was defined as infected or not infected at the time of dry- off based on one or more quarters being infected. Clustering of cows within herds was

78 accounted for by using generalized estimating equations (GEE) to calculate sensitivity and specificity of each definition explained above. A range of positive and negative predictive values were generated for a series of prevalence levels using three (B, C and E) of the definitions created.

4 A Results

4.4.1. Samples

Eight hundred and twenty quarter samples (from 206 cows) with complete records

(gold standard culture, somatic cell count, three Petrifilm cultures and Tri-plate culture) were obtained. Ninety-four samples (11.5%) were contaminated (three or more colony types) on gold standard and thus were not included in the analysis. After removal of the contaminated samples, there were still 206 cows in the study, however, only 137 cows had samples from all 4 quarters included. Analysis was conducted with and without accounting for clustering within cows. When GEE were used to account for cows clustered within herds, estimates for

Se and Sp of each definition changed very little (<1.0%); however, confidence intervals were narrowed slightly.

4.4.2. Gold Standard

The prevalence of pathogens based on microbiological culture of quarter samples taken within one week of dry-off was 43.1%. Figure 4.1 shows the results of gold standard culture, based on the percentage of quarter samples infected with a particular organism. Both primary and secondary microorganism (if present) isolated are shown. The category ‘Other

Gram Positive’ includesBacillus spp., and the category ‘Other Gram Negative’ includes

Serratia spp., Proteus spp., Pseudomonas spp. andPasteurella multocida. The most common

79 primary organism isolated was CNS from 195 quarter samples (26.9%), followed by ‘Other

Gram Positive’ from 60 quarter samples (8.3%), Strep, spp. from 24 quarter samples (3.3%) andS. aureus from 13 quarter samples (1.8%). The remaining organisms recorded each accounted for less than one percent of the samples received. Very few samples (7.6%) had a second microorganism isolated. In most cases, the second organism isolated was an ‘Other

Gram Positive’ (3.3%) or a CNS (2.8%).

When the NMC criteria of significance were applied to the results, only 11.9%

(86/726) of quarter samples were classified as having microorganisms of probable or high significance. This was because many samples received cultured fewer than 10 colonies of

CNS which is not considered significant unless it is a pure growth of more than 10 colonies.

After data were collapsed to the cow level, 33.0% (68/206) of cows were considered positive on gold standard culture, meaning they had at least one quarter that was positive for a significant mastitis-causing organism.

4.4.3. 3M Petrifilm

The prevalence of quarter samples that were diagnosed positive (5 or more colonies) using the AC Petrifilm was 39.0%. When data were aggregated to the cow level, 59.2%

(122/206) cows were considered infected. When the AC Petrifilm was used alone to classify a cow positive for significant bacterial growth it was highly sensitive, 100% (95% Cl: 94.7 -

100) and moderately specific, 61.2% (95% Cl: 53.8 -68.1).

A total of 34 CC Petrifilms had colony growth from individual quarter samples; however, only 2 had 20 or more colonies present, therefore test characteristics were not calculated for this test.

80 There were 13 quarter samples submitted that culturedS. aureus on gold standard culture. As with the Petrifilm CC plate, cow level results were not computed. All of the 13 samples were found to be positive forS. aureus using the STX Petrifilm and confirmatory

STX disk, resulting in a sensitivity of 100.0%. Specificity was slightly 95.5% because 32 of the 713 samples that were notS. aureus on the gold standard test grew pink colonies on the

STX disk. Most (27) of these samples cultured CNS on gold standard.

Using the Petrifilm STX and the presence of red-violet colonies alone (i.e. no disk applied) to diagnose a sample as positive forS. aureus also resulted in a sensitivity of

100.0%. However, the specificity of the test when used in this manner was very poor, at

27.9%. This was because there were 214 plates that exhibited red-violet colonies that were not diagnosedS. aureus on gold standard. The gold standard culture results for these 214 samples were: 107 CNS, 75 no growth, 18 other gram positive, 9 Streptococcus spp, 2

Corynebacterium spp., 2 S. hyicus and 1 unidentified gram negative organism.

4.4.4. Minnesota Easy Culture System II Tri-plate

The prevalence of quarter samples that resulted in colony growth (at least one colony on any section of the Tri-plate) was 31.0%. After aggregating the data, 68.9% (142/206) of cows were classified as infected. Using the Tri-plate alone to classify a cow positive for significant bacterial growth resulted in a sensitivity of 96.4% (95% Cl: 90.2 - 98.8) and a specificity of 44.6% (95% Cl: 35.9 -53.6). The lower specificity was the result of 76 cows that the Tri-plate diagnosed positive that the gold standard did not.

Any growth on the Factor portion of the Tri-plate was further examined for the presence of hemolysis. As with the STX Petrifilm, this analysis was completed using quarter

81 samples only. When hemolysis alone was used to diagnose a sample as positive forS. aureus, the sensitivity was 76.9% (95% Cl: 46.2 - 95.0) and the specificity was 84.5% (95%

Cl: 78.6-89.3).

4.4.5. Somatic Cell Count

Somatic cell counts were analyzed at two different thresholds, 150 000 and

200 000. Composite cell counts from the DHI test just prior (within a month) of dry-off ranged from 10 000 to 11 658 000, with a mean of 293 000 and a median of 104 000. Table

4.2 shows the calculated sensitivities and specificities of the two definitions used to consider a cow infected, based on SCC alone.

4.4.6. Combination of Somatic Cell Count and Culture

Sensitivity and specificity for each definition was calculated using the NMC definition for significant growth to classify cows as positive on gold standard. Table 4.2 shows the results of these calculations for the two cow level somatic cell count thresholds that were examined. The highest sensitivity (100%) was achieved using the Petrifilm alone to classify a sample as positive or negative for colony growth. Therefore, combining the

Petrifilm culture result with somatic cell count results in parallel (i.e., positive on one test or the other) could not result in an increase in sensitivity. Using the Tri-plate alone also resulted in a high sensitivity (96.4%). But unlike the Petrifilm, when combined with somatic cell count results (positive on one test or the other), sensitivity of the test was modestly increased.

Positive and negative predictive values for a range (0-1) of disease prevalences were examined for three of the definitions: cow somatic cell count >150 on any of the last three

82 tests, positive on AC Petrifilm (5 or more colonies) and positive on Tri-plate (at least one colony on any section of the plate). These results are presented graphically in Figures 4.2 and

4.3. For prevalences between 0 and 50%, the PPV ranged from 0 to 61 % and NPV ranged from 100 to 68%.

4.5. Discussion

The aim of the present study was to determine the ability of the University of

Minnesota Tri-plate, the 3M Petrifilm and somatic cell count data to classify cows as infected or uninfected at the time of dry-off. Previous studies have evaluated the use of somatic cell count data (6) and CMT results (8) at the time of dry-off to determine the infection status of the cow. This study is unique in that it evaluates two recently validated culture based tests that may be used on-farm to screen cows before subjecting them to a selective dry cow therapy program. Laboratory and field evaluations have determined that both the 3M

Petrifilm and the University of Minnesota Tri-plate are successfully able to identify groups of mastitis causing pathogens from clinically affected cows (11, 15). Selective dry cow therapy has been extensively investigated and in previous studies, quarters that were not treated with antibiotics consistently had higher rates of IMI and clinical mastitis, both during and after the dry period (18,19,20). In Sanfordet al. (7) it was concluded that non-antibiotic dry cow therapy can be as effective as antibiotic treatment if used in uninfected quarters. For this reason, diagnostic tests used to select cows that should receive antimicrobials at the time of dry-off must have maximal sensitivity in order to minimize the proportion of infected mammary quarters that would go untreated.

83 The samples used in this study were obtained from cows on dairy farms in the

Atlantic Canadian provinces, thus the prevalence of pathogens are representative of this region. In our study, the prevalence of quarters infected with any pathogen was43.1% using the gold standard microbiology procedures. Previous studies have reported prevalences of infection at dry-off between 28% to 50% at the cow level (6). This variation in prevalence may be due in part to the differences in study design and gold standard definitions the authors used for infection (6). Most recently Torreset al. completed a study in 4 Ohio dairy herds reporting an overall prevalence of infection of 34.3% based on single quarter samples. In that study, a quarter was considered infected if >100 cfu/ml of major contagious pathogens

(S. aureus or S. agalactiae) or >500 cfu/ml of any other pathogens were isolated (6). In the present study, the NMC guidelines of significance were applied to the quarter samples and only 11.9% were found to be ‘significantly’ infected. In 2006, Sanfordet al. completed a study in Eastern Canada and found that 16% of quarter samples were infected with either a major pathogen or a minor pathogen with >10 colony forming units per 0.1ml of milk of either CNS or C. bovis (8). As in the Torres and Sanford studies, the most common pathogen isolated in this study was CNS (26.7% of quarter samples). When the significance guidelines were applied, only 8% of quarter samples were considered infected with CNS, a minor mastitis-causing pathogen. It is of interest that previous work has questioned the value of treating quarters with minor IMIs. Some studies have demonstrated that IMI with these organisms decreased the risk of developing IMI caused by major pathogens (21). However, there have been other studies have reported different results. In an invited review,

DeVliegher et al. reported that the description of pathogenicity of CNS varied from being

84 protective, to being indifferent to udder health, to being the cause of subclinical infection and even the cause of mild clinical mastitis (22). The diagnostic tests used in this study do not have the ability to differentiate CNS from other gram positive mastitis-causing pathogens, therefore, if they were to be used as part of a selective dry cow therapy program, any cow with a CNS infection at the time of dry-off would be considered infected and receive antimicrobials.

The 3M Petrifilms (AC and CC) used in this study have been previously evaluated for their ability to classify clinical infections as being caused by either gram positive or gram negative pathogens (12). In the present study, very few (<1.0%) of the quarter samples were positive for a Gram negative pathogen on gold standard culture, therefore the test characteristics of the CC Petrifilm were not evaluated. The percentage of cows classified as infected using the Petrifilm was higher than that of the gold standard (59.2% and 33.0% respectively); this could be the result of the larger innoculum of milk that was used (100 pL vs. lOpL on the gold standard cultures). In previous work, the cut-off point of 5 colonies was chosen to maximize the sensitivity of the AC Petrifilm realizing that false positive test results would occur. Clinically, the consequences of treating a false positive are less of a concern than not treating an animal that is truly infected.

Recent work indicates that these diagnostic tests may not necessarily be producing many false positive results. The NMC Guidelines traditionally used create a pool of gold standard positives that can be considered “strong positives”, however, recent work suggests that many samples may in fact be false negatives on gold standard cultures. Dohooet al. found that when wanting to identify as many existing infections as possible, then the criteria

85 for considering a quarter sample positive should be a single colony isolated from a 0.01 mL sample of milk (23).

The Petrifilm STX plate was also used in this study to diagnose quarter samples with

S. aureus. The STX has been previously evaluated using milk samples from clinically infected cows (15,24,25). Very few (13) samples were positive for S. aureus on gold standard culture, all of which were confirmed positive using the Petrifilm STX, resulting in a sensitivity of 100%. There were 32 samples that were diagnosedS. aureus using the

Petrifilm (confirmed with the STX disk) that were not classified as such using the gold standard. As in Chapter 3, the application of the STX disk did not affect the sensitivity of the test as the product information suggests, but did improve its specificity (from 27.9% to

95.5%).

In our study, colonies that grew on the STX plate were not subcultured to confirm their identity, however in Chapter 3, where pink colonies were subcultured, it was found that

45% (5/11) were in fact S. aureus that had not been identified by the original culture. As with the AC Petrifilm, it may have been possible to detect moreS. aureus on the STX

Petrifilm because a larger innoculum of milk was used.

The University of Minnesota Tri-plate has also been previously evaluated for its ability to classify pathogens from clinically infected cows (12). As the focus of this study was to identify a test that could be used to detect an infection at dry-off, the Tri-plate was not fully evaluated for its ability to differentiate pathogens. When used to detect an infection caused by any pathogen, the resulting sensitivity and specificity were 96.4 and 44.6%, respectively. Because S. aureus is a major contagious pathogen that can be very difficult to

8 6 cure, the Tri-plate’s ability to detect it was also evaluated with a resulting sample level sensitivity and specificity of 76.9% and 84.5%, respectively. In Chapter 3, we found a Se and

Sp of 97.9% and 81.8%, respectively when the Tri-plate was used to diagnoseS. aureus on clinical mastitis samples (15). The difference in Se and Sp from this work and Chapter 3 may be attributed to decreased sheddingS. ofaureus from non-clinical pre dry-off samples or the technician reading the Tri-plate. However, the relevance of apparently large disagreement in sensitivities of the test is questionable due to the difference in sample size between the two studies. In Chapter 3,46/47 samples were correctly identified S.as aureus, while in the current chapter, there were 10/13 correctly identified. Since the diagnosisS. ofaureus was based solely on the presence of hemolysis around colonies on the Factor medium, the readers’ ability to detect hemolysis influenced the performance of the test. This was identified in a previous study, where the sensitivity of the Tri-plate ranged from 43.2 - 59.1% and the specificity ranged from 93.8 - 95.9% when read by readers with limited microbiology training (15).

Somatic cell count data are readily accessible to producers that are enrolled in monthly DHI testing. In 2008, Torreset al. investigated the use of clinical mastitis history in combination with SCC data to identify cows with infection at the time of dry-off. They found that infected and uninfected cows at dry-off were most efficiently identified using three months SCC records with a threshold of 200 000 for cows without clinical mastitis during the lactation and a threshold of 100 000 for cows that had a case of clinical mastitis in the first 90 days of lactation. Using those criteria, the efficiency [(number of true positives)

+ (number of true negatives)/ total number of samples] of the test was 64.8%, the sensitivity

87 was 69.7% and the specificity was 62.4% (6). Previous work has repeatedly shown that a cut-off of approximately 200 000 to 250 000 cells/mL was optimal to reduce diagnostic error when distinguishing between infected and uninfected quarters (5). Our study evaluated two

SCC cut-offs (150 000 and 200 000) as well as single point-in-time samples (last sample prior to dry-off) and data from 3 months prior to dry-off to classify cows as infected. As with previous work, our study found that using data from 3 months prior to dry-off resulted in a higher sensitivity (42.6%) than data from just one test prior to dry-off which gave a sensitivity of 36.9%. As was expected, the sensitivity of the test was also improved using the lower cell count threshold but was still not considered high enough to be used to guide treatment decisions at dry-off. A similar study by Middletonet al. published in 2004 concluded that SCC data were not sensitive enough to be used as a screening test for identifying infected quarters in a herd with a high bulk tank SCC (26).

Our study also examined various ways to combine data from somatic cell count records with results from culture based tests. Results from the quarter cell counts were combined with those of both the Petrifilm and Tri-plate at both cut-offs of SCC (data not presented). Interpreting these tests in series resulted in lower sensitivities (between 43.0% and 54.7%) and higher specificities (between 88.0% and 92.5%), as expected. When designing a testing method to identify cows that are to be subjected to a selective dry cow therapy program, it is important to maximize the sensitivity of the test. When multiple tests are interpreted in parallel, the sensitivity will be improved. In the present study, it was not possible to improve the sensitivity (100%) of the Petrifilm by combining it with somatic cell count data.

88 The positive and negative predictive values of each of the individual tests were evaluated as they represent the ability of a test to predict disease. Of particular interest is the

negative predictive value, as a producer that is making a decision to withhold treatment based

on a negative test, wants to be confident in the test result. The predictive value of a test is

influenced by the prevalence of disease in the population (26). Therefore, predictive values

were calculated across a range of possible prevalences. The proportion of cows that test

negative that are not truly negative can be calculated (1-NPV). In herds with a low

prevalence of mastitis infection (i.e. between 10-15%), this proportion would be very small,

<1% for both the Petrifilm and Tri-plate, and <8% for 3 months SCC data at a threshold of

150 000. These results indicate that very few cows that were infected at the time of dry-off

would go without antimicrobials in a selective dry cow therapy program.

4.6. Conclusions

Both the 3M Petrifilm and University of Minnesota Tri-plate are rapid and reliable

tests that could be used successfully in a selective dry cow therapy program. Using the 3M

AC Petrifilm alone to classify a sample as infected, resulted in a sensitivity of 100% and a

specificity of 61.2%. When Petrifilm results were interpreted in parallel with SCC data no

improvement in test sensitivity was possible but decreases in specificity were noted. Using

the Tri-plate alone resulted in a sensitivity of 96.1% and a specificity of 44.6%. When Tri­

plate results were combined with SCC data, the sensitivity of the test was improved to

98.5%, however, the specificity of the test was decreased to 31.7% and 34.2% using 150 and

200 000 as SCC cut offs, respectively. These results suggest that both the 3M Petrifilm and

89 University of Minnesota Tri-plate are sensitive tests that could be used as part of selective dry cow treatment program.

4.7. Acknowledgments

This work was supported by Canadian Bovine Mastitis Research Network and

Maritime Quality Milk at the Atlantic Veterinary College. We thank to the many staff and students at the AVC who were involved in sample collection, Doris Poole, Shana Richard,

Zoe Little and Jane Saunders.

4.8. References

1. 3M Microbiology. 2005. 3M Petrifilm Interpretation Guide. 3M Microbiology, Saint Paul, MN.

2. Laboratory for Udder Health. 2000. Minnesota Easy Culture System II Handbook. Minnesota Veterinary Diagnostic Laboratory, University of Minnesota, Saint Paul, MN.

3. Bradley, A. J. and M. J. Green. The importance of the nonlactating period in the epidemiology of intramammary infection and strategies for prevention. 2004. Vet. Clin. North Am. Food Anim. Pract. 20:547-68.

4. Berry, E. A. and J. E. Hillerton. 2002. The effect of selective dry cow treatment on new intramammary infections. J. Dairy Sci. 85:112-21.

5. Schukken, Y.H., D. J. Wilson, F. Welcome, L. Garrison-Tikofsky and R. N. Gonzalez. 2003. Monitoring udder health and milk quality using somatic cell counts. Vet. Res. 34(5):579-96.

6. Torres A. H., P. J. Rajala-Schultz, F. J. Degraves and K. H. Hoblet. 2008. Using dairy herd improvement records and clinical mastitis history to identify subclinical mastitis infections at dry-off. J. Dairy Res. 75:240-7.

7. Sanford C.J., G. P. Keefe, I. R. Dohoo, K.E. Leslie, R. T. Dingwell, L. DesCoteaux and H. W. Barkema. 2006. Efficacy of using an internal teat sealer to prevent new intramammary infections in nonlactating dairy cattle. J. Am. Vet. Med. Assoc. 228:1565-73.

90 8. Sanford C.J., G. P. Keefe, J. Sanchez, R. T. Dingwell, H. W. Barkema, K. E. Leslie and I. R. Dohoo. 2006. Test characteristics from latent-class models of the California Mastitis Test. Prev. Vet. Med. 77:96-108.

9. Sargeant J. M., K. E. Leslie, J. E. Shirley, B. J. Pulkrabek and G. H. Lim. 2001. Sensitivity and specificity of somatic cell count and California mastitis test for identifying intramammary infection in early lactation. J. Dairy Sci. 84:2018-24.

10. Hochalter J., S. Godden, R. Bey, A. Lago and M. Jones. 2006. Validation of the Minnesota easy culture system II: Results from on-farm bi-plate culture versus standard laboratory culture. Page 250-251 in Am. Assoc. Bovine Pract. Proc. St. Paul, MN.

11. Neeser, N. L., W. D. Hueston, S. M. Godden and R. F. Bey. 2006. Evaluation of the use of an on-farm system for bacteriologic culture of milk from cows with low-grade mastitis. J. Am. Vet. Med. Assoc. 228:254-260.

12. McCarron, J. L., G. P. Keefe, S. L. McKenna, I. R. Dohoo and D. E. Poole. 2009. Laboratory evaluation of 3M Petrifilms and University of Minnesota Bi-plates as potential on-farm tests for clinical mastitis. J. Dairy Sci. 92:2297-2305.

13. National Mastitis Council. 1999. Laboratory and Field Handbook on Bovine Mastitis. 2nd ed. Natl. Mastitis Counc. Inc., Madison, WI.

14. National Mastitis Council. 1987. Laboratory and Field Handbook on Bovine Mastitis. 1st ed. W. D. Hoard and Sons Co., Fort Atkinson, WI.

15. McCarron, J. L., G. P. Keefe, S. L. McKenna, I. R. Dohoo and D. E. Poole. 2009. Evaluation of the University of Minnesota Tri-plate and 3M Petrifilm for the isolation of Staphylococcus aureus and Streptococcus species from clinically mastitic milk samples. J. Dairy Sci. 92:5326-33.

16. Reyher, K. K., S. Dufour, H. W. Barkema, L. Des Coteaus, T. J. DeVries, I. R. Dohoo, G. P. Keefe, J. P. Roy, and D. T. Scholl. 2011. The National Cohort of Dairy Farms - A data collection platform for mastitis research in Canada. J. Dairy Sci. 94: 1616-1626

17. StataCorp. 2005. Intercooled Stata 9.0. College Station, Texas, USA.

18. Robinson, T. C., E. R. Jackson and A. Marr. 1983. Within herd comparison of teat dipping and dry cow therapy with only selective dry cow therapy in six herds. Vet. Rec. 112:315-9.

91 19. Schukken, Y. H., J. Vanvliet, D. Vandegeer and F. J. Grommers. 1993. A randomized blind trial on dry cow antibiotic infusion in a low somatic cell count herd. J. Dairy Sci. 76:2925-30.

20. Berry, E. A. and J. E. Hillerton. 2002. The effect of an intramammary teat seal on new intramammary infections. J. Dairy Sci. 85:2512-20.

21. Lam, T. J., Y. H. Schukken, J. H. van Vliet, F. J. Grommers, M. J. Tielen and A. Brand. 1997. Effect of natural infection with minor pathogens on susceptibility to natural infection with major pathogens in the bovine mammary gland. Am. J. Vet. Res. 58:17- 22.

22. DeVliegher, S., L. K. Fox, S. Piepers, S. McDougall and H. W. Barkema. 2012. Invited Review: Mastitis in dairy heifers: nature of the disease, potential impact, prevention and control. J. Dairy Sci. 95: 1025-1040.

23. Dohoo I.R., J. Smith, S. Andersen, D.F. Kelton, S. Godden, and Mastitis Research Workers’ Conference. Diagnosing intramammary infections: Evaluation of definitions based on a single milk sample. 2011. J. Dairy Sci. 94:250-261.

24. Silva, B. O., D. Z. Caraviello, A. C. Rodrigues and P. L. Ruegg. 2005. Evaluation of Petrifilm for the isolation of staphylococcus aureus from milk samples. J. Dairy Sci. 88:3000-3008.

25. Wallace, J., J. Roy, E. Bouchard, L. DesCoteaux, S. Messier and D. DuTremblay. 2008.Comparison of 3M Petrifilm Staph Express Count plates, 3M Petrifilm Rapid Coliform Count plates and 3M Aerobic Count plates with standard bacteriology of bovine milk. Page 162-163 in Natl. Mast. Counc. Ann. Mtg., New Orleans, LA.

26. Middleton J. R., D. Hardin, B. Steevens, R. Randle and J. W. Tyler. 2004. Use of somatic cell counts and California mastitis test results from individual quarter milk samples to detect subclinical intramammary infection in dairy cattle from a herd with a high bulk tank somatic cell count. J. Am. Vet. Med. Assoc. 224:419-23.

27. Dohoo, I., W. Martin and H. Stryhn. 2009. Veterinary Epidemiologic Research. 2nd ed. VER Inc., Charlottetown, PE, Canada.

92 Table 4.1. Criteria used to create definitions for test positive samples positive test for definitions create to used Criteria 4.1. Table KQ>Titnonw> Definition ______Cow SCC >200 or 150 on any of the last three tests previous to dry-off dry-off to tests previous three last the of 150any on or >200 SCC Cow Cow SCC >200 or 150 on last test previous to dry-off dry-off to previous test 150last on or >200 SCC Cow Cow SCC >200 or 150 on any of the last three tests or positive on Tri-plate on positive tests or three last the of any 150 on or >200 SCC Cow Petrifilm on positive tests or three last the of any 150 on or >200 SCC Cow Tri-plate on positive or test last 150 on or >200 SCC Cow plate) the of section any on colony one least (at on Tri-plate Positive positive Petrifilm or test last 150 on or >200 SCC Cow colonies) more or (5 Petrifilm AC on Positive 93 Table 4.2. Cow level sensitivity and specificity (95% Cl computed by GEE) of all test definitions evaluated versus gold standard status based on NMC 1987 significance

Sample Culture tests only Definition Sensitivity Specificity C 100 (94.7 -100) 61.2 (53.8-68.1) E 96.4(90.2 - 98.8) 44.6 (35.9 - 53.6) DHI SCC only Sensitivity 150 Specificity 150 Sensitivity 200 Specificity 200 A 50.2(39.1- 61.3) 75.9 (69.8-81.1) 36.9 (28.7-46.0) 82.1 (74.0-88.0) B 64.9(46.7 - 75.0) 67.7(60.8 - 74.0) 42.6 (35.3 - 50.3) 78.1 (68.5-85.5) DHI SCC and cuture combination Sensitivity 150 Specificity 150 Sensitivity 200 Specificity 200 D 100 (94.7 - 100) 43.2 (37.5-49.1) 100 (94.7 - 100) 48.5(41.3-55.7) F 98.5(90.8-99.8) 31.7 (35.9-38.5) 96.4(90.2-98.8) 34.2 (27.0 - 42.1) G 100 (94.7 - 100) 36.9 (31.1-43.2) 100 (94.7 - 100) 43.8 (37.1-50.8) H 98.5 (90.8-99.8) 26.3 (20.1-33.5) 96.4 (90.2 - 98.8) 30.4 (23.7 - 38.3)

A - Cow SCC >150 or 200 on last test B - Cow SCC >150 or 200 on any of the last three tests C - Positive on AC Petrifilm (5 or more colonies) D - Cow SCC >150 or 200 on last test or Petrifilm positive E - Positive on Tri-plate (at least one colony on any section of the plate) F - Cow SCC >150 or 200 on last test or positive on Tri-plate G - Cow SCC >150 or 200 on any of the last three tests or positive on Petrifilm H - Cow SCC >150 or 200 on any of the last three tests or positive on Tri-plate

94 iue41 ure odsadr utr eut o 2 uresfo 0 fcw n 18 on cows of from 206 quarters 726 for results culture standard gold Quarter 4.1. Figure

Percent 00 - 40.0 n 60.0 50.0 20.0 - 30.0 10.0 0.0 - farms. Microorganismlsolated on Gold Standard Gold Standard on Microorganismlsolated a. Q. ■ Primary■ Species ■ Secondary Species Secondary ■ Species 95 JO a- B- oo Culture

o. q . as

© .05 ,45 v Definition B Definition C Definition E

Figure 4.2. Cow level positive predictive values for a range of disease positive prevalence

96 a> co

o .05 .35 .45 Prevalence Definition B ------Definition C Definition E

Figure 4.3. Cow level negative predictive values for a range of disease positive prevalence

97 CHAPTERS. SUMMARY AND GENERAL DISCUSSION

98 5.1. Introduction

The overall objective of this study was to complete a thorough laboratory validation of potential on-farm mastitis pathogen identification systems and investigate ways they could be applied in the field.

To accomplish this objective, the tests were first used to place samples of clinically mastitic milk into two treatment categories (Chapter 2). The Minnesota Easy Culture System

II and the 3M Petrifilm were successful at categorizing samples into the broad categories of gram positive, gram negative or no growth. The University of Minnesota Bi-plate had been specifically designed for such use; however, this study was the first to evaluate the use of the

3M Petrifilm in such a manner.

Both test systems had the ability to further differentiate major mastitis-causing pathogens from clinically affected cows by using the University of Minnesota Tri-plate and the 3M Staph Express. In addition to therapy decisions, important management decisions can be made when a producer knows the pathogen causing an infection. Therefore, a second study was carried out to validate the use of the tests in this manner (Chapter 3).

Finally, because the tests proved successful at categorizing cases of clinical mastitis that would benefit from lactational therapy, it was felt that they could also be useful in a decision model to target therapy at the time of dry-off. Dry cow antimicrobial therapy accounts for a large proportion of antibiotics used on dairy farms. Selective dry cow therapy programs are continuing to gain interest, however, to implement such programs, reliable tests are needed that can determine the infection status of a cow at dry-off and thus guide therapy

99 decisions. Therefore, the tests’ ability to identify the infection status of cows at dry-off was also examined (Chapter 4).

5.2. Minnesota Easy Culture System n

Previous studies have validated the use of University of Minnesota Culture System as a tool to guide therapy decisions (1,2). However, none of the studies that validated the test had been published in peer reviewed journals. The test system is commercially available and in use on many farms in the US. In 2006, Neeseret al. published a study that looked at farms using the Bi-plate and concluded that on-farm systems for bacteiiologic culture of milk may result in significant reductions in the percentage of cows treated with antimicrobials (3).

Since the ultimate goal of this study was to identify a test that would have the same end result on Canadian dairy farms, a thorough laboratory evaluation was required before considering it for use in the field.

The gold standard used for the evaluation of this test (as well as others throughout the study) was microbiologic culture with the NMC guidelines of significance applied. These guidelines place results in categories of significance not simply infected or not. By using the probable and highly significant categories to define a sample as gold standard positive, the gold standard may in fact have limited sensitivity for gram positive infections. This is supported by recent work by Dohoo et al. which found that when identifying as many existing infections as possible is important, then the criteria for identifying a sample as positive is the presence of a single colony from a 0.01 mL milk sample (4).

Using an imperfect gold standard, as was done in these studies can result in biases.

100 Samples from truly uninfected cows would be unlikely to meet the guidelines of significance, therefore, there were probably few false positives on gold standard. However it is possible that some samples that were truly infected did not meet the NMC criteria, thus creating false negative gold standard results. The consequences of this misclassification include 1) Some of the gold standard negative results would be truly positive. If these samples had tested positive on any of the tests evaluated in this study, they would have been considered false positives and potentially negatively biased the Sp estimates presented. 2) Conversely, the samples that the evaluated tests classified positive would have been easier to diagnose using a gold standard that identifies “strongly positive” cases. The effect of this would be an upward bias in Se (4). These same biases would be present in all of the test evaluations in this paper.

5.2.1 Minnesota Easy Culture System II Bi-plate

The attributes of the Bi-plate that were of interest were its diagnostic characteristics

(sensitivity and negative predictive value especially), the percentage of samples that the Bi­ plate would categorize as treatment candidates and finally, it ease of use which was evaluated by comparing agreement between blinded readers. The Bi-plate was highly sensitive

(97.9%) in its ability to determine treatment categories for clinical cases of mastitis. The negative predictive value was also very high (96.4%) in the study population, which had a disease prevalence of 54.6% using the gold standard technique. In Chapter 2, a single prevalence from a nationwide study completed by Olde Riekerink (5) was used to demonstrate that the negative predictive value remained high (98.5%). When the Bi-plate was used to categorize samples into two treatment categories, 63.8% were classified as

101 treatment candidates (positive for gram positive growth). Therefore, using a Bi-plate culture to determine which cases of clinical mastitis receive antimicrobial therapy could potentially result in a 36.2% reduction in antimicrobial treatment, if the current practice of a producer was to treat all clinical cases (Chapter 2). To simulate the ease of use in a field condition, Bi­ plates were read by a microbiology laboratory technician and 4 blinded readers with limited training. The agreement between readers was very good with a combined kappa statistic of

0.76. The Bi-plate performed very well in this laboratory evaluation. In addition to the laboratory results, other practical aspects of using the test must be considered. In this case, the most important practical implications are the shelf life and requirement for strict refrigerated conditions in transport and storage. Since the media used in the Bi-plate is blood-based the shelf life is limited to approximately 6 weeks. This could prove to be problematic for smaller dairy farms that would only require tests sporadically.

5.2.2 Minnesota Easy Culture System II Tri-plate

The Tri-plate provides the same diagnostic capabilities as the Bi-plate, in that it contains the same media that allow differentiation between gram negative and gram positive pathogens (Factor and MacConkey). Additionally it contains MTKT medium, which is specific for Streptococcus spp. The Factor medium that is used in the Tri-plate is a blood based agar which will show P-hemolysis in the presence ofS. aureus. In standard laboratory diagnostics, the identification of p-hemolysis alone is not considered a reliable method for the diagnosis ofS. aureus in milk; however some research has been done to test its reliability.

In 1995, Lam et al. reported that 20 to 25% of the S. aureus isolates from bovine mastitis did not present detectable p-hemolysis in primary cultures (6. In the laboratory evaluation of the

102 Tri-plate, its sensitivity was very high when the laboratory technician used it to diagnose S. aureus based solely on P-hemolysis, however, the inter-reader agreement was poor when used by the blinded readers with limited training (Chapter 3). In contrast, the specificity of the test was higher for the inexperienced readers, ranging from 93.8 to 95.9%, possibly because these readers were more reluctant to call a colony positive. There may be some merit to using the Tri-plate in the above manner if the users of the test are properly trained, however caution should be used when important management decisions are being made based on the diagnosis ofS. aureus from a Tri-plate.

The Tri-plate also contains a third section of medium that is specific for the growth of

Streptococcus species (Modified TKT). When read by the technologist, both the Se and Sp of the test were high, 92.6 and 89.5%, respectively (Chapter 3). The Tri-plate has the ability to be used in the same manner as the Bi-plate to categorize cases into treatment categories, to determine infection status at the time of diy-off (Chapter 4) and to provide further diagnostic information in which management decisions may be based on. It would be a useful test for producers wishing to obtain as much diagnostic information as possible using a single sample on-farm. However, as with the Bi-plate, the same practical concerns with shelf life and storage conditions exist.

5 3 .3M Petrifilm

3M Petrifilms are sample ready plates that were designed to be used commercially for rapid bacteriological isolation and enumeration of bacteria from food products (7). Their use for the detection of mastitis causing pathogens in whole milk was first reported in 2004 by

Silva et al. in a study that focused primarily on the diagnostic capabilities of the Staph

103 Express Plate (8). The use of the Aerobic count plate and Coliform count plates was mentioned briefly as part of an on-farm treatment protocol; however, their diagnostic capabilities were not reported.

5.3.1. Aerobic Count and Coliform Count Petrifilms

At the beginning of the study, whole milk samples were used as per the manufacturer’s instructions (7), however, the Petrifilm was not designed for use with mastitic milk. Milk clots and cases of heavy bacterial growth resulted in difficulty identifying individual colonies; therefore a second set of Petrifilms was evaluated using milk samples that had been diluted 1:10. When the Petrifilms were used with diluted milk samples, the sensitivity and negative predictive values were not different from undiluted samples (Chapter

2). There were differences in specificity (undiluted 72.6 vs. diluted 92.9%) and positive predictive values (undiluted 77.7 vs. diluted 93.8%), both of which favored using the diluted milk samples. For this reason, further analysis to choose appropriate colony count thresholds for the Petrifilms was done using only the diluted milk samples.

Since this was the first study to use the Aerobic count plate and Coliform count plate in combination to create treatment categories, appropriate cut off values had to be created as to when to consider a plate positive for bacterial growth. A range of cut points were evaluated using sensitivity versus specificity plots that were created for various combinations of colony count thresholds ranging from 1 to 30, in increments of 5. The combination of

Petrifilms that resulted in the optimum sensitivity (93.8%) used a threshold of <20 colonies on the Coliform count Petrifilm and >5 colonies on the Aerobic count Petrifilm to consider the test positive (Chapter 2).

104 5.3.2. Staph Express Petrifilm

The Staph Express Plate has been evaluated in previous studies for its ability to diagnoseS. aureus from milk samples (9,10). The test is designed to initially detect

Staphylococcus species and with the addition of a confirmatory disk and further incubation, identifyS. aureus. The manufacturer’s instructions state that the presence of red-violet colonies after the initial culture is indicative ofS. aureus, but if the colour is not easily distinguished or other colony colours exist, then the Staph Express disk can be added to confirm its presence. The disk contains a dye and deoxyribonucleic acid.S. aureus produces deoxyribonuclease (DNase) and the DNase reacts with the dye to form confirmatory pink zones on the Petrifilm (7). As the addition of a disk and re-incubation of the plate may be cumbersome in on-farm situations, Chapter 3 looked at both ways to use the test, diagnosing

S. aureus with and without the disk. The Staph Express Petrifilm had a higher sensitivity when using only the presence of red-violet colonies than when the presence of pink colonies on the Staph Express Disk was used, 97.4 versus 92.1%. The specificity of the test was greatly improved by the use of the disk, from 76.1 to 93.1%.

5.4. Use in Selective Dry Cow Therapy

In Chapter 4, the University of Minnesota Tri-plate and the 3M Petrifilms were used to determine infection status of a cow at the time of dry-off. The Tri-plate was chosen for this study as it provided the most diagnostic information of the two University of Minnesota tests; however, for simply determining infection status of a cow, the Bi-plate could have been used in its place. All three of the Petrifilms were used but since very few samples were positive for Coliforms, the Coliform Count plate could not be properly evaluated in the study.

105 The Aerobic Count plate alone provided enough diagnostic information as to the infection status of the cow, so it was the focus of the analysis.

Previous work has evaluated cow side tests for determining infection status at dry-off

(11,12), however a highly sensitive test has not been identified. Both the Tri-plate and

Petrifilm were evaluated alone and in combination with somatic cell count results obtained from DHI records. The Tri-plate and Petrifilm were successful at detecting infected cows with sensitivities of 96.1% and 100% respectively (Chapter 4). When Petrifilm results were interpreted in parallel with SCC data it was not possible to improve the test sensitivity.

When Tri-plate results were combined with SCC data, the sensitivity of the test was improved slightly.

55. Overall Conclusions

The objectives of this project were met by evaluating a variety of microbiologic tests in a laboratory setting. Tests were evaluated on their abilities to place samples into treatment categories, identify specific species of interest and to determine infection status at dry-off.

The University of Minnesota Bi-plate and the 3M Petrifilm were successfully able to categorize clinical cases of mastitis into two treatment categories based on their ability to detect the presence of a gram positive organism. The University of Minnesota Tri-plate and the 3M STX Petrifilm were able to successfully detectS. aureus in clinical milk samples and the 3M Petrifilm and University of Minnesota Tri-plate proved that they could be used successfully in a selective dry cow therapy program.

106 5.6. Future Directions

All of the tests that were evaluated in this study performed very well in laboratory conditions. The next logical step in their evaluation is to have them used on-farms by dairy producers. To evaluate their use to target therapy during lactation, a large scale field study will incorporate results of the on-farm test into a treatment algorithm and follow up to determine the effects on antibiotic use, disease outcomes (bacteriologic and clinical cure, culling and relapse) and overall incidence of clinical mastitis on the farm. The expectation is that the use of an on-farm test will reduce the amount of antibiotics used for treating and preventing mastitis without compromising the health of the animal, however this must be thoroughly evaluated in the field. Other aspects of on-farm culture that should be looked at are the practical factors that would affect the outcome of their performance including the quality of samples that are used, the conditions that the tests are used in, shelf life of the test and the ability of the user to read the test and interpret the results. To be considered successful, the test should have the ability to provide an economic benefit for the producer by decreasing the amount of milk that is discarded from treated animals and reducing the amount of money that is spent on treatments. A similar field study should also be completed to evaluate the utility of a cow side test as part of a selective dry cow therapy program.

Despite extensive research in the field of mastitis, it still remains a very costly disease for dairy producers. For now, producers can make the best of the current situation by arming themselves with the most information possible and making wise evidence based decisions when it comes to mastitis therapy and prevention. On-farm diagnostic tests may offer

107 producers the opportunity to make those evidence based decisions and will hopefully become common place on Canadian dairy farms in the near future.

108 5.7. References

1. Hochhalter, J., S. Godden, R. Bey, A. Lago and M. Jones. 2006. Validation of the Minnesota Easy Culture System II: Results from in-lab bi-plate culture versus standard laboratory culture, and bi-plate inter-reader agreement. Page 298 in 39th Annu. Proc. Am. Assoc. Bovine Pract., Saint Paul, MN.

2. Lago, A., S. Godden, R. Bey, K. Leslie, R. Dingwell and P. Ruegg. 2006. Validation of the Minnesota Easy Culture System II: Results from on-farm bi-plate culture versus standard laboratory culture. Page 250-251 in Am. Assoc.Bov. Prac. Ann. Mtg. Proc., St.Paul, MN.

3. Neeser, N. L., W. D. Hueston, S. M. Godden and R. F. Bey. 2006. Evaluation of the use of an on-farm system for bacteriologic culture of milk from cows with low-grade mastitis. J. Am. Vet. Med. Assoc. 228:254-260. 4. Dohoo I.R., J. Smith, S. Andersen, D.F. Kelton, S. Godden, and Mastitis Research Workers’ Conference. Diagnosing intramammary infections: Evaluation of definitions based on a single milk sample. 2011. J. Dairy Sci. 94:250-261.

5. Olde Riekerink, R., H. Barkema, D. Kelton and D. Scholl. 2008. Incidence rate of clinical mastitis on Canadian dairy farms. J. Dairy Sci. 91:1366-1377.

6. Lam, T. J. G. M., A. Pengov, Y. H. Schukken, J. A. H. Smit and A. Brand. 1995. The differentiation of Staphylococcus aureus from other micrococcaceae isolated from bovine mammary glands. J. App. Bacteriol. 79:69-72. Paul, MN.

7. 3M Microbiology. 2005.3M Petrifilm Interpretation Guide. 3M Microbiology, Saint Paul, MN.

8. Silva, B., D. Caraviello, A. Rodrigues and P. Ruegg. 2004. Use of petrifilm for mastitis diagnosis and treatment protocols. Page 52 in Natl. Mastitis Counc. Ann. Mtg. Proc., Charlotte, NC.

9. Silva, B. O., D. Z. Caraviello, A. C. Rodrigues and P. L. Ruegg. 2005. Evaluation of Petrifilm for the isolation of staphylococcus aureus from milk samples. J. Dairy Sci. 88:3000-3008.

10. Wallace, J., J. Roy, E. Bouchard, L. DesCoteaux, S. Messier and D. DuTremblay. 2008. Comparison of 3M Petrifilm Staph Express Count plates, 3M Petrifilm Rapid Coliform Count plates and 3M Aerobic Count plates with standard bacteriology of bovine milk. Page 162-163 in Natl. Mast. Counc. Ann. Mtg. Proc., New Orleans, LA.

109 11. Sanford C.J., G. P. Keefe, J. Sanchez, R. T. Dingwell, H. W. Barkema, K. E. Leslie and I. R. Dohoo. 2006. Test characteristics from latent-class models of the California Mastitis Test. Prev. Vet. Med. 77(l-2):96-108.

12. Torres A. H., P. J. Rajala-Schultz, F. J. Degraves and K. H. Hoblet. 2008. Using dairy herd improvement records and clinical mastitis history to identify subclinical mastitis infections at dry-off. J. Dairy Res. 75(2):240-7.

110 Appendix 1

NMC Guidelines for significance of colony numbers of specific organisms isolated pure or with other colony types (based on 0.01 ml quarter sample)

Total Number Colonies One Several (2-10) More than 10 Culture Pure Pure Mixed two types Mixed Several Types Pure Mixed two types Mixed Several Types

S. agalaciae 44 4 44 4 4 Group G Strep 4 4 4 4 4 4 4 Streptococcal Spp 2 3 2 2 4 3 1 S. aureus 3 4 4 4 4 4 4 Staphylococcal Spp (CNS) 1 2 2 2 4 2 1 E. coli 2 3 2 2 4 2 1 Klebsiella 2 3 2 2 4 2 1 Enterobacter 2 3 2 2 4 2 1 Serratia 2 3 2 2 4 2 1 Pasteurella 4 4 4 4 4 4 4 Pseudomonas 2 3 2 2 4 4 2 Yeast, Mold, Fungi 2 3 1 1 4 2 1 Nocardia 2 3 2 2 4 3 3 Prototheca 2 3 3 2 4 3 3 C. bovis 1 2 2 2 4 3 3 A. pyogenes 2 3 3 3 4 3 3 C. ulcerans 2 4 3 2 4 4 3 Proteus 2 3 1 1 4 2 1

Degree of confidence in diagnosing an infection: 1 - not significant 2 - questionable significance 3 - probable significance 4 - highly significant

111